Hallucinations can significantly undermine end-user trust. They arise from various factors, including:

• Patchy, insufficient or false training data. It results in the Large Language Model (LLM or model) fabricating information when it’s unsure of the correct answer.
• Model lacks proper grounding and context to determine factual inaccuracies.
• Excessive model complexity for the application.
• Inadequate software testing.
• Poorly crafted, imprecise or vague end-user prompts.

Organizations can mitigate the risk and frequency of these hallucinations occurring. That avoids embarrassing the company and misleading its customers by adopting multiple strategies, including:

• Clear model goal.
• Balanced training data.
• Accurate training data.
• Adversarial fortification.
• Sufficient model tuning.
• Limit responses.
• Comprehensive model testing.
• Precision prompts.
• Fact-check outputs.
• Human oversight.

Let’s explore the last five of these mitigations in more detail. To read about the first five, click here.

Limit responses

Models produce hallucinations more often when they lack constraints that limit the scope of possible outputs. To improve the overall accuracy of outputs, define boundaries for models using filtering tools, maximum word lengths and clear probabilistic thresholds for the acceptability of outputs. These limits reduce the risk of hallucinations.

For example, when the model cannot assign a sufficient confidence level to a proposed recommendation about optimizing a production process, it should not provide that output to an engineer.

Comprehensive model testing

Inadequately tested models produce more hallucinations than comprehensively tested models.

Testing typically detects hallucinations by cross-referencing model-generated output with other trusted and authoritative sources.

It’s easy to recommend testing models rigorously before production use. It is vital to preventing or at least dramatically reducing the risk of hallucinations. However, software development teams are always under schedule pressure, and testing is the easiest task to shortchange because it occurs near the end of the project.

For example, project managers must assertively remind management of the costs and reputational risks of releasing inadequately tested models for production use.

Precision prompts

Ambiguity or lack of specificity in prompts can result in the model generating hallucinations or output that doesn’t align with the end-user’s intent. That result decreases confidence in the model or causes misinterpretation or misinformation.

Asking the right question is essential to achieve superior outputs from models. Accurate, relevant outputs depend on the clarity and specificity of engineers’ prompts. Precision prompts that reduce hallucinations exhibit these features:

• Maximize clarity and specificity by writing prompts that are as short and specific as possible.
• Provide context such as time, location or unique identifiers to narrow the scope.
• Use descriptive language by specifying relevant characteristics such as profession, discipline, industry or geographic region.
• Plan an iterative approach by refining successive prompts based on previous outputs.
• Minimize the risk of biased outputs by ensuring fairness and inclusivity.

For example, write a specific prompt like “How is consistency achieved in stamping steel automotive wheels?” Avoid a general prompt like “How is quality achieved in manufacturing wheels?”

Fact-check outputs

Sometimes, hallucinations are not recognized and used by engineers in their work with dangerous or expensive consequences.

Engineers can reduce this hallucination risk by:

• Fact-checking the output against other sources.
• Asking the model to describe its reasoning and data sources.
• Checking if the output is logically consistent and aligns with general world knowledge.
• Writing a slightly different prompt to see if the model produces the same output.

For example, a prompt about a chemical additive should not produce output about a closely related but materially different chemical.

Human oversight

Once a model is in routine production use, it’s tempting for engineers to move on to address the next AI opportunity. However, not monitoring the performance of your AI application means you have no sense of the:

• Number of hallucinations it’s producing.
• Need to adjust or retrain the model as data ages and evolves.
• Evolving end-user requirements that need to be addressed through model enhancements.

A better practice is to assign an analyst to regularly sample model outputs to validate their accuracy and relevance. Analysts can spot hallucinations that suggest model refinement is necessary.

For example, a model designed to support problem diagnosis for complex production machinery may occasionally provide an inaccurate investigation recommendation.

By implementing these best practices, engineers can significantly reduce model hallucinations and build confidence in the reliability of model outputs to advance their digital transformation.

The post How can engineers reduce AI model hallucinations – part 2 appeared first on Engineering.com.

Offsetting all this considerable value, generative AI sometimes produces inaccurate, biased or nonsensical output that appears authentic. Such outputs are called hallucinations. The following types of hallucinations occur:

• Output doesn’t match what is known to be accurate or true.
• Output is not related to the end-user prompt.
• Output is internally inconsistent or contains contradictions.

Hallucinations can significantly undermine end-user trust. They arise from various factors, including:

• Patchy, insufficient or false training data. It results in the Large Language Model (LLM or model) fabricating information when it’s unsure of the correct answer.
• Model lacks proper grounding and context to determine factual inaccuracies.
• Excessive model complexity for the application.
• Inadequate software testing.
• Poorly crafted, imprecise or vague end-user prompts.

AI hallucinations can have significant consequences for real-world applications that erode engineers’ confidence. For example, an AI hallucination can:

• Provide inaccurate values leading to an erroneous engineering load calculation and product failure.
• Suggest stock trades leading to financial losses.
• Incorrectly identify a benign skin lesion as malignant, leading to unnecessary medical interventions.
• Contribute to the spread of misinformation.
• Inappropriately deny credit or employment.

Organizations can mitigate the risk and frequency of these hallucinations occurring. That avoids embarrassing the company and misleading its customers by adopting multiple strategies, including:

• Clear model goal.
• Balanced training data.
• Accurate training data.
• Adversarial fortification.
• Sufficient model tuning.
• Limit responses.
• Comprehensive model testing.
• Precision prompts.
• Fact-check outputs.
• Human oversight.

Let’s explore the first five of these mitigations in more detail.

Clear model goal

Hallucinations occur if the model goal is too general, vague, or confusing. An unclear model goal will make the selection of appropriate training data ambiguous. That leads to an increase in the frequency of hallucinations.

A small amount of team collaboration can often clarify the model goal and reduce hallucinations.

For example, a model goal to verify machine performance is too general. A better goal might be to verify steel lathe or stamp performance.

Balanced training data

Hallucinations occur if the training data used to develop the model is insufficient, unbalanced or includes significant gaps. That leads to edge cases where the model attempts to respond to prompts with inadequate data. Overfitting is the term used to describe a model trained on a limited dataset that can’t make accurate predictions.

Train your model on diverse, representative data that covers a wide range of real-world examples for your application domain. Ensuring your training data is representative may require creating synthetic data. Use a data template to ensure all training data instances conform to a standard data structure. That improves the quality of training data and reduces hallucinations.

For example, in responding to a prompt about copper’s tensile stress, a model that contains primarily data about steel and aluminum performance characteristics will not have sufficient exposure to other metals.

Accurate training data

Many models are trained on data read from many public web pages. Many problems cause inaccurate web information that can lead to hallucinations, including:

• Simple spelling and grammatical errors or misunderstandings.
• Deliberately vague or erroneous information designed to mislead people.
• Information that was correct in the past but has been superseded by updates or new research.
• Humour, irony or parody that is easily misunderstood.
• Contradictory information due to conflicting opinions or scientific theories.
• Errors introduced by translating information from another language.

You can’t fact-check every web page you’ve used to build training data. However, you can fact-check a sample of web pages to estimate the risk of errors and related hallucinations in your training data.

For example, an engineer may prompt a model to verify a calculation. If the second output is different, the engineer may be able to identify inaccurate training data.

Adversarial fortification

Adversarial attacks consist of prompts intentionally or unintentionally designed to:

• Launch a cyber attack to create financial loss, brand reputation damage, or intellectual property theft.
• Mislead the model to produce hallucinations to compromise the reliability and trustworthiness of the model.

A model can become more resistant to adversarial attacks by:

• Integrating adversarial examples into the training process to improve the classifier’s resistance to attack.
• Introducing algorithms designed to identify and filter out adversarial examples.
• Including adversarial examples in the scope of model testing.

For example, an engineer might unintentionally write a prompt that outputs a hallucination. The engineer should report the output to the team managing the model. The team should implement an enhancement that will reduce the likelihood of future hallucinations.

Sufficient model tuning

Hallucinations increase if a model is inadequately tuned.

Model tuning is a manual and semi-automated experimental process of finding the optimal values for hyperparameters to maximize model performance and reduce hallucinations. Hyperparameters are variables whose values the model cannot estimate from the training data.

For example, in responding to an engineer’s prompt about wind tunnel performance, a model that has not been inadequately tuned may return values that violate the laws of fluid behaviour.

By implementing these best practices, engineers can significantly reduce model hallucinations and become more confident in the reliability of model outputs.

The post How can engineers reduce AI model hallucinations? appeared first on Engineering.com.

• Working as an engineering assistant.
• Engaging in more wide-ranging problem-solving with less effort and elapsed time.
• Automating business workflows more quickly and at a lower cost.
• Developing software more rapidly and with less effort.

Many engineers are walking along their personal digital transformation path. Sensors and software have replaced printed log books for data acquisition. Online systems have replaced file folders for data management. Excel has replaced paper worksheets. Digital displays have replaced large corkboards or walls decorated with many sticky notes. Data analytics became feasible once companies transformed most of their data into digital datastores.

These examples of how engineers use generative AI to drive value from digital transformation should trigger specific ideas for you to pursue.

Generative AI as an engineering assistant

Increasingly capable computers and improved application software have digitally transformed the work of engineers. Engineers access a myriad of data from many sources. It needs to be integrated, cleansed and analyzed to create value from all this mostly digital data. That work can be tedious and error-prone.

Generative AI has opened up new possibilities for engineers to view AI as an automated engineering assistant. Now, engineers can look more widely for data and simulate more possible solutions while reducing effort and the risk of errors.

“Strive to treat AI as an assistant and not simply as a new tool,” said Mel Head, a retired chemical engineer with a computing background who worked at Honeywell. “We still need to review the AI output before using it exactly as we would with a human assistant.”

On-target search

Once upon a time, engineers relied on printed reference texts for much of their work. Engineers have relied on Google’s digital search results like everyone else for many years. Search has brought much of the information engineers need for their work to their fingertips. When did engineers last visit a reference library or buy an engineering textbook?

Generative AI offers the ability to avoid the effort of sifting through Google’s digital search results by summarizing the available information and focusing on what’s most important. That AI ability assures engineers that nothing significant has been missed while reducing their effort.

“Using GPT-4o or Perplexity for search has pretty much supplanted Google searches for me. Rather than Google something and get a million links and many ads to sort through, GPT-4o and Perplexity give me well-formed and cited answers, said Jeff Uhlich, Principal Consultant at Obleeq Solutions. “They’re not always 100% accurate, but they are more useful than Google results.”

Powerful data analytics

Engineers have always analyzed data. First, slide rules, then calculators and then Excel. Each represented a significant advance for engineers.

In digitally transformed businesses, generative AI can analyze and summarize vast volumes of data to produce meaningful tables and impressive charts. That AI functionality improves the quality of reports while reducing engineers’ effort.

“I use Microsoft Data Analysis Expressions (DAX) and Power Query M to analyze oil and natural gas production data. AI tools like OpenAI ChatGPT and Microsoft CoPilot provide code snippets and links to reference material to confirm syntax details,” said Mark Perrin, petroleum engineer and VP at TriAcc Group. “The bottom line is that these AI tools save me a lot of time when solving client data problems.”

More capable business workflows

Every business operates with many workflows. Engineers often play a significant role in adapting workflows to changing business conditions or capturing the value of new technology.

Digital transformation creates an opportunity to incorporate more digital data into automated workflows. Then, generative AI can add more sophisticated decision-making to these workflows.

“With platforms like Zapier and Make to automate my workflows, AirTable to quickly develop low-code apps, and Notion to integrate diverse office data, I’ve streamlined and automated a plethora of business operations,” says Alan Mourgues, consulting reservoir engineer and founder at CrowdField. “From generating content for blogs and websites to data analysis, our business processes are fluid, responsive to business changes and efficient.

Faster software development

Developers and engineers have handcrafted digital software since the invention of computers. This work has been expensive, tedious and error-prone.

More recently, sophisticated integrated software development environments (IDE) and the wide availability of function-rich open-source software libraries have greatly improved developer productivity.

The advent of generative AI has improved developer productivity further by generating significant amounts of source code based on a comparatively short prompt. These advances will help software developers meet the voracious software appetite of engineers and our society more generally.

“ChatGPT is a powerful tool for engineers to develop software rapidly,” says Damien Hocking, CTO at Madala Software. “AI tools don’t give you a perfect solution, but it’s fantastic as a flying start that saves hours of grunt work.”

Digital research organization

Engineers are constantly involved in improving production performance and designing new products. That work includes significant research into new materials, automation and process advances. Digitally organizing the results of this work has been difficult. Document management systems and apps like Microsoft Notes have helped. When project teams collaborate, apps like Slack and Microsoft Teams preserve the discussion well.

Generative AI adds the ability to summarize and prioritize research results digitally. That saves engineers vast amounts of reading time.

“I’ve started replacing my extensive paper notes with a digital workflow on a tablet using the Nebo app,” said Brad Henrie, quality and process engineer at Sealweld Corporation. “My notes combine sketches, flowcharts, text, and calculations, so I can’t achieve the same results with a word processor. Nebo performs simple arithmetic, converts my handwritten text to computer-readable text and is searchable.”

Generative AI increases engineers’ productivity in a digitally transformed business without compromising quality or adding risk.

The post Enhancing your personal digital transformation path appeared first on Engineering.com.

Too often, engineers find they’ve added managing their project sponsor to that lengthy list. That perspective is misguided. Instead, engineers should view their project sponsor as someone who can help them, someone who can be expected to accept specific thorny tasks that only a senior executive can deftly handle for the benefit of the project.

Here are some of the ways that project sponsors help engineers. Consider using some of the points in this article during your next meeting with your project sponsor.

You can explore these and other tips to help project sponsors and engineers be more effective in our new book, A Project Sponsor’s Warp-Speed Guide – Improving Project Performance. It’s available from Amazon at this link.

Communicates project benefits throughout the organization

The project sponsor and steering committee members must enthusiastically communicate, sell and defend the project benefits in meetings and informal discussions throughout the organization. If these individuals fail to champion the benefits or, worse, challenge the benefits or criticize the project, the project is doomed.

For example, these leaders frequently remind the organization of the project’s value proposition to maintain its commitment to the project at various management meetings or town hall events.

When this visible public support is not evident, engineers provide project sponsors and steering committee members with brief talking points to encourage more communication.

Guides the project manager

The project sponsor guides the project manager. The project sponsor offers organizational insights about internal politics, corporate history, and prejudices held by various stakeholders. This information is valuable to engineers, who often do not have enough seniority and reputation for the organization to accept their necessary but unwelcome recommendations.

When engineers feel neglected, they can reach out to their project sponsors to reconfirm the following best practice points for building their relationship:

• Commit to a firm schedule of meetings with the project manager. The frequency is usually weekly or bi-weekly.
• Respect the project manager’s mandate and delegation.
• Provide open, frank feedback to the project manager on project observations and what improvements are needed.
• Demand honest opinions from the project manager about project status and issues.
• Will not create pressure to provide a false, overly optimistic project status.
• Operate the project manager relationship based on mutual trust.

Conversely, if the project sponsor loses confidence in the engineer, the project sponsor must replace them.

Encourages the team

The project sponsor occasionally speaks to the entire team to publicly provide kudos, express encouragement and boost morale. On these occasions, the project sponsor strongly supports the project and the team’s work.

For example, the project sponsor can share some senior management scuttlebutt and organization performance metrics that would be good for the team to hear and reinforce the importance of the team’s work for the organization.

When teams feel unappreciated, engineers can diplomatically encourage their project sponsors to inspire the team.

Ensures resource commitments are fulfilled

When the project was approved, various stakeholders accepted resource commitments to work with the project. However, as the project proceeds, the business departments are typically hit with new resource demands and gradually de-commit from the project. Only the project sponsor can reverse this trend.

For example, only the project sponsor can effectively glare at senior managers or VPs to rebuild the commitment. Engineers can’t do that and survive in the organization.

It’s up to engineers to point out this failure to fulfill commitments to their project sponsors for resolution.

Resolves issues that the project manager cannot resolve

Every project develops issues related to scope, priorities and approach. Only the project sponsor can resolve or lead the resolution of the more significant issues that tend to cross organizational lines.

For example, the project depends on manufacturing data, and the data quality is low. Only the project sponsor can march into the office of the VP of Manufacturing and ask that the data be cleaned up and extract a commitment that the data will remain high quality into the future.

It’s up to engineers to raise these issues with their project sponsors for resolution.

Shields the team from distracting internal politics

To the greatest extent possible, the project sponsor shields the team from distracting and harmful internal politics. The project sponsor also defends the project team from being hijacked to solve a crisis in the business.

For example, if the team is distracted and upset by rumours of a reorganization or downsizing, the project sponsor can reassure the team.

It’s up to engineers to raise the concerns with their project sponsors for attention.

You can explore these and other tips for effective project sponsors in A Project Sponsor’s Warp-Speed Guide – Improving Project Performance, a new book I wrote with my co-author Jocelyn Lapointe. It’s available from Amazon at this link. View the book as a reference tool. You don’t have to read it all to obtain actionable insights.

The post How project sponsors help project managers succeed appeared first on Engineering.com.

Scoping a digital transformation program is never easy because the number of attractive opportunities available to engineers almost always exceeds the available budget and staff capacity. Sometimes, ambitious blue sky projects are approved because their vision is so appealing. That never ends well and wastes significant resources. Every project should be scoped to advance some aspect of the business strategy.

Here’s a five-step process for scoping projects that will avoid boondoggles and deliver value for engineers and their organizations.

Prioritize opportunities

Before engineers can start work on any digital transformation project, they need to understand which opportunities offer the highest return and lowest risk. Use these questions to rank proposed projects:

• Does the scope of the proposed projects address significant pain points and opportunities? This question differentiates urgent needs from wants.
• Are there senior executives prepared to be the project sponsors and champions of the projects? Proposals without a project sponsor should be rejected due to likely failure.
• How appealing are the business cases for the proposed projects? Low-return proposals should receive a low priority.
• How does the scope of the proposed projects align with the business strategy? Poor alignments indicate lower-priority proposals.
• Are any of the proposed projects prerequisites to other projects? Sometimes, foundational projects, often with little or no business case, must be completed before other higher-benefit projects can start.
• How risky is the scope of the proposed projects? Projects requiring extensive data cleanup, emerging technologies or significant specialty skills should be deferred.
• How enormous is the scope of the proposed projects? Projects that are too small don’t advance digital transformation much. Projects requiring massive budgets or organizational resources should be split up or rejected.
• Will the scope of any of the proposed projects take more than one year to complete? Longer projects consume significant staff resources and are at higher risk of not finishing. These should be broken up or deferred.

Answering these questions leads to a ranked list of digital transformation projects and rejected proposals. Projects with the highest return and lowest risk will be near the top. Engineering managers can then approve as many projects as budget and staff constraints allow.

Understand the business context

Before engineers can scope an approved digital transformation project, they need to understand the business context and the drivers for change. Start by answering these questions and documenting the answers in the project charter:

• Who are the key stakeholders, such as internal departments, customers, and vendors involved or affected by the project?
• What are the needs, expectations, and preferences of the key stakeholders?
• How will the project scope impact the business processes, culture, and organization structure?
• What external forces, such as competition, technological advances or political change, are driving the project scope?

If you can’t answer some of these questions, you must conduct more analysis or conclude that the opportunity is not as appealing as initially thought.

Write the project charter

The project charter is a document that summarizes the project goal, supporting objectives, deliverables, scope, business case, organization, budget, assumptions, constraints, and risks. It serves as a basis for planning, executing, monitoring, and controlling the project. Writing this document requires the collaboration of the project sponsor, team members and stakeholders to reach a consensus on every element. Consensus-building helps avoid misunderstandings, disputes, or rework later during the digital transformation project.

If the collaborating groups can’t reach a consensus on the project characteristics, such as scope, one or more of the following issues are occurring:

• Lack of agreement on scope and priorities.
• Wishful thinking about what the allocated resources can achieve.
• Personality or power conflicts.
• A fantasy business case.
• A focus on technology potential rather than business value.
• An attempt to tackle more scope than the organization can absorb.

The organization may be unprepared for the opportunity if the project manager and project sponsor can’t facilitate the needed consensus. It’s best to stop before the project starts and consumes significant resources for no business beefit.

Build support for the project scope

Once the project charter has been accepted, it’s time to build support for the digital transformation project scope among a wider audience of senior management, stakeholders, and employees. Helpful techniques used to review and validate the project scope include:

• Scope feedback – a continuous process of collecting and analyzing feedback from customers, end-users, or other stakeholders on the project deliverables or outcomes).
• Scope validation – an informal process of testing, inspecting, or evaluating the project goal, objectives, scope description and deliverable list to confirm they meet the scope and high-level requirements.
• Scope verification – a more formal process of obtaining acceptance and sign-off of the project charter and its scope description.

If the project manager and project sponsor encounter significant challenges or hesitancy while building support, the project charter may need to be revised. If stakeholders challenge the business case or the scope description, the opportunity is not as appealing as initially thought. Then, it’s best to stop before the project starts.

Manage scope changes

Once a project is underway, scope change proposals are unavoidable as the understanding of project details grows. Because digital transformation projects involve exploration, uncertainty, and innovation, the frequency of scope change proposals will be higher than for other project types.

When stakeholders see that you’re managing a project effectively, they want more scope. However, you won’t be a hero if you deliver more scope later than the original project completion date. Engineers can respond to this push for more scope as follows:

• Asking stakeholders to fund a change order only to analyze the impact of their proposed scope changes. If they don’t support that, you’ve dodged a bullet.
• If your team produces an analysis report of the impact of their scope changes, recommend that the work be deferred to the likely follow-on project.
• If your stakeholders want the change now, ask them what other scope they want to remove from the project to maintain the budget and schedule. Since the stakeholders want everything in the current scope, the new idea is dropped.

Once the project team recognizes that you’re not approving any proposed scope changes, they will question why the proposals are being documented and sometimes analyzed. The value of the documentation is that the unapproved scope changes provide an excellent starting point for scoping a follow-on project.

When engineers manage digital transformation projects by including only the scope described in the project charter and avoid adding appealing new scope, they will deliver successful projects.

The post Managing digital transformation scope appeared first on Engineering.com.

Many engineers are investigating digital transformation initiatives using artificial intelligence (AI) features within their organizations.

They’re generally aware of various AI risks and are searching for ways to better categorize, understand, mitigate and communicate them.

A group of scientists from the Massachusetts Institute of Technology (MIT) and other universities are also concerned with AI risks, so they built an AI Risk Repository to serve as a common frame of reference for discussing and managing AI risks.

They classified AI risks into seven AI risk domains:

1. Discrimination & toxicity.
2. Privacy and security.
3. Misinformation.
4. Malicious actors and misuse.
5. Human-computer interaction.
6. Socioeconomic and environmental.
7. AI system safety, failures and limitations.

When engineers use these AI risk domains to conduct risk management for their digital transformation initiatives, they will gain a better understanding of how AI can help their digital transformation initiatives without causing irreparable harm.

Discrimination & toxicity

The discrimination and toxicity AI risk domain consists of the following subdomains:

1. Unfair discrimination and misrepresentation – Unequal treatment of individuals or groups by AI.
2. Exposure to toxic content – AI systems expose end-users to harmful, abusive, unsafe, or inappropriate content.
3. Unequal performance across groups – The accuracy and effectiveness of AI decisions and actions are directly related to group membership.

If this risk becomes a reality, the business impact includes loss of reputation and the distraction and cost of lawsuits. If this risk turns into reality, the societal implications include unfair outcomes for discriminated groups.

Privacy and security

The privacy and security AI risk domain consists of the following subdomains:

1. Compromise of privacy – AI systems obtain, leak, or correctly infer sensitive information.
2. Security attacks – AI systems exploit systems, software development tool chains and hardware vulnerabilities.

If this risk turns into reality, the business impact includes data and privacy breaches and loss of confidential intellectual property, leading to regulatory fines and the cost of lawsuits. If this risk becomes a reality, the societal implications include:

• Compromising end-user privacy expectations, assisting identity theft, or causing loss of confidential intellectual property.
• Breaches of personal data and privacy.
• System manipulation causing unsafe outputs or behavior.

Misinformation

The misinformation AI risk domain consists of the following subdomains:

1. False or misleading information – AI systems inadvertently generate or spread incorrect or deceptive information.
2. Pollution of the information ecosystem and loss of consensus reality – Highly personalized AI-generated misinformation creates “filter bubbles” where individuals only see what matches their existing beliefs.

If this risk turns into reality, the business impact includes misleading performance information, employee mistrust and potentially dangerous product designs. If this risk becomes a reality, the societal implications include inaccurate beliefs in end-users that weaken social cohesion and political processes.

Malicious actors and misuse

The malicious actors and misuse AI risk domain consists of the following subdomains:

1. Disinformation, surveillance and influence at scale – Using AI systems to conduct large-scale disinformation campaigns, malicious surveillance, or targeted and sophisticated automated censorship and propaganda.
2. Cyberattacks, weapon development or use and mass harm – Using AI systems to develop cyberweapons, develop new or enhance existing weapons.
3. Fraud, scams and targeted manipulation – Using AI systems to gain a personal advantage over others.

If this risk becomes a reality, the business impact includes loss of business continuity, cost to recover from attacks and bankruptcy. If this risk turns into reality, the societal implications include:

• Manipulating political processes, public opinion and behavior.
• Using weapons to cause mass harm.
• Enabling cheating, fraud, scams, or blackmail.

Human-computer interaction

The human-computer interaction AI risk domain consists of the following subdomains:

1. Overreliance and unsafe use – End-users anthropomorphizing, trusting, or relying on AI systems.
2. Loss of human agency and autonomy – End-users delegate critical decisions to AI systems, or AI systems make decisions that diminish human control and autonomy.

If this risk turns into reality, the business impact includes misleading and potentially dangerous system outputs. If this risk becomes a reality, the societal implications include compromising personal autonomy and weakening social ties.

Socioeconomic and environmental

The socioeconomic and environmental AI risk domain consists of the following subdomains:

1. Power centralization and unfair distribution of benefits – AI-driven concentration of power and resources within certain entities or groups.
2. Increased inequality and decline in employment quality – The widespread use of AI leads to social and economic disparities.
3. Economic and cultural devaluation of human effort – AI systems capable of creating economic or cultural value reproduce human innovation or creativity.
4. Competitive dynamics – Competition by AI developers or state-like actors perpetuates an AI “race” by rapidly developing, deploying and applying AI systems to maximize strategic or economic advantage.
5. Governance failure – Inadequate regulatory frameworks and oversight mechanisms fail to keep pace with AI development.
6. Environmental harm – The development and operation of AI systems cause environmental damage, partially through their enormous electricity consumption.

If this risk becomes a reality, many businesses will likely collapse due to economic collapse. If this risk turns into reality, the societal implications include:

• Inequitable distribution of benefits and increased societal inequality.
• Destabilizing economic and social systems that rely on human effort.
• Reduced appreciation for human skills.
• Release of unsafe and error-prone systems.
• Ineffective governance of AI systems.

AI system safety, failures and limitations

The AI system safety, failures, & limitations AI risk domain consists of the following subdomains:

1. AI pursuing its own goals in conflict with human goals or values – AI systems act in conflict with ethical standards or human goals or values.
2. AI possessing dangerous capabilities –  AI systems develop, access, or are provided with capabilities that increase their potential to cause mass harm through deception, weapons development and acquisition, persuasion and manipulation, political strategy,cyber-offense, AI development, situational awareness and self-proliferation.
3. Lack of capability or robustness – AI systems fail to perform reliably or effectively under varying conditions, exposing them to errors and failures that can have significant consequences.
4. Lack of transparency or interpretability – Challenges in understanding or explaining the decision-making processes of AI systems.
5. AI welfare and rights – Ethical considerations regarding the treatment of potentially sentient AI entities.

If this risk becomes a reality, the business impact includes loss of reputation and the distraction and cost of lawsuits. If this risk turns into reality, the societal implications include:

• AI using dangerous capabilities such as manipulation, deception, or situational awareness to seek power or self-proliferate.
• Mistrusting AI systems.
• Enforcing compliance standards becomes difficult.

Do not expect every digital transformation initiative to have risks in every subdomain. Nonetheless, there is value in explicitly considering every subdomain during the risk identification task.

When digital transformation teams identify and mitigate project risks using these AI risk domains, they will ensure their risk management processes are as comprehensive as possible.

The post Managing AI risks in digital transformation appeared first on Engineering.com.

The power of artificial intelligence (AI) to enhance digital transformation initiatives has become increasingly evident to engineers as they seek to improve operational efficiencies, scale innovation and gain a competitive edge.

While digital transformation is hardly new, AI and large language models (LLMs) have emerged as a formidable accelerator by changing business processes, reshaping products and services and sometimes upending entire industries.

AI’s ability to learn and improve over time, coupled with digital transformation, means that organizations can realize faster processes, reduced costs and more efficient operations.

These AI benefits contribute to an environment of continuous improvement and innovation that is often key to success in a competitive environment.

Enhancing data-driven decision-making

Engineers know that data-driven organizations use digital insights to shape strategies, optimize processes and respond rapidly to market changes. However, harnessing the potential of integrated digital data at scale for decision-making requires far more than traditional data analytics.

AI’s capability adds unprecedented speed and precision to decision-making by:

• Sifting through structured data, identifying patterns and generating predictive insights.
• Analyzing vast volumes of unstructured data more effectively than search engines, specialized databases or software developers to generate predictive insights.
• Avoiding the cost and elapsed time associated with custom data integration of diverse data sources using software developers.
• Autonomously detecting trends and forecasting outcomes.

Adding AI and LLM capability to data-driven decision-making helps engineers optimize operational and strategic decisions while reducing the need to base decisions on history, experience, in-vogue ideas or hunches.

Examples of adding AI capability to data-driven decision-making for engineering include:

• Monitoring large volumes of IIoT data from production equipment to identify performance anomalies to avoid unscheduled downtime.
• Sifting through the external media for general and industry audiences to identify competitor initiatives that may require a response.
• Summarizing patent data, trademark data and research journals maintained in multiple languages to identify potentially relevant technology developments.

Automating processes and workflows

Automation is a fundamental aspect of digital transformation. AI-powered tools like robotic process automation (RPA), machine learning and cognitive computing, a type of AI that simulates human thought processes, have taken digital transformation to new heights.

While valuable, previous generations of automation that engineers implemented were limited to well-defined, repetitive tasks and detailed, rule-based decisions. AI expands automation to more complex decision-making processes, pattern recognition and more generalized problem-solving.

Examples of adding AI capability to automating processes and workflows include:

• Adding more accuracy and sophistication to simulations. For example, engineers can refine and enhance their designs through successive simulations to reduce limitations, which leads to more innovative solutions.
• Enhancing supply chain management for better product demand forecasting, logistics optimization, order fulfillment and risk assessments for component shortages. Achieving these improvements requires the integration of disparate data sources maintained by partners.
• Adding more intelligence to RPA transaction workflows such as invoice and shipment receipt processing. Examples include identifying potentially duplicate invoices, assessing the materiality of discrepancies and identifying likely fraud.

Improving data quality

Engineers are painfully aware that insufficient data quality is the number one reason for the failure of digital transformation initiatives. Asking data analysts to identify and correct data quality issues is slow, tedious, expensive and subject to further errors.

AI can automate data quality improvement work using pattern recognition. AI can achieve more speed and consistency at a lower cost than human analysts.

Examples of using AI capability to automate data quality improvement include:

• Recognizing that existing equipment can’t manufacture the designs due to dimensions, lack of accessibility and unachievable tolerances.
• Identifying and correcting instances where numeric values are associated with different units of measure or measurement systems creates errors and confusion.
• Sharply reducing the number of duplicate and incomplete inventory master records.
• Generating synthetic data to augment existing datasets to improve AI models.

Persisting knowledge

Organizations lose surprising amounts of essential knowledge and intellectual property (IP). Too often, engineers reinvestigate problems or wrestle again with design refinements because of a lack of awareness of prior work. Loss of knowledge and expertise typically occurs due to:

• Staff turnover and transfers.
• Reluctance to share knowledge.
• Lack of management support for knowledge management.
• Lack of time to document work.
• No repository in which to store work products.
• No easy ability to search and retrieve documents.
• Organization restructuring, acquisitions and mergers.
• Confusion caused by inaccurate, outdated or redundant versions of information.

Addressing these issues without digital transformation is impossible. Including digital knowledge management to the scope of digital transformation initiatives can significantly increase the value organizations achieve from the knowledge they have accumulated, often at considerable effort and cost.

Adding AI agents to knowledge repositories can add another increment of value. AI agents are intelligent software that use an LLM to perform query tasks, make decisions and learn from their experiences like humans. AI agents are a significant advance on the more familiar chatbots.

Examples of using AI agents to enhance digital knowledge management include enabling engineers to:

• Query “tribal knowledge” to improve production performance.
• Discover best practices.
• Better troubleshoot production equipment problems based on records of historical incidents.
• Query IP such as patent records, test results, research reports, development studies and licensing agreements in support of current work.

Challenges and considerations

While AI and LLMs add potential to digital transformation, engineers must acknowledge the challenges and ethical considerations associated with its deployment. These include:

• Ensuring data privacy to maintain customer and employee confidence.
• Addressing biases in AI algorithms and training data to maintain trust and inclusivity.
• Recognizing that LLMs may be incomplete or misleading.
• Training a skilled workforce that can effectively manage AI-driven processes.
• Fostering a culture that embraces innovation to ensure the smooth integration of AI and LLM technologies.

Engineers can establish robust data governance, prioritize transparency in communication and continuously monitor AI systems to mitigate unintended consequences.

Artificial intelligence is a powerful accelerator of digital transformation. Its impact spans most industries and functions, enhancing efficiency, agility and resilience. By embracing AI’s transformative potential, businesses can achieve a sustainable competitive advantage and drive long-term growth.

The post Where AI can accelerate digital transformation appeared first on Engineering.com.

Digital transformation reshapes many industries by changing how organizations operate and deliver customer value. Many engineers have recognized the compelling benefits of digital transformation, including increased productivity. Through advanced software, automation, data analytics, and enhanced connectivity, digital transformation enables businesses to operate more efficiently, innovate faster, and deliver better outcomes.

Below, we explore more ways in which digital transformation boosts productivity. To read the first article in this series, click here.

Reduce costs and optimize resources

Digital transformation often leads to significant cost savings by reducing waste, improving asset utilization, and optimizing resource allocation. Through digital technologies like the Industrial Internet of Things (IIoT), simulation and digital twins, engineers can monitor equipment and processes in real-time, ensuring that resources like energy and materials are used more efficiently.

Predictive maintenance, enabled by IIoT sensors and AI, helps businesses reduce downtime and prevent costly repairs by identifying issues before they become critical. Avoiding unscheduled outages contributes significantly to production productivity. Cloud service providers (CSP) can reduce IT infrastructure costs, which have become a material expense, by allowing businesses to:

• Pay for only the CSP resources they use.
• Avoid investing in an on-premise computing environment.
• Avoid operating costs for an on-premise computing environment.
• Utilize a CSP computing environment with superior cybersecurity defenses.
• Access enormous CSP computing resources instantly when needed.

Leading simulation software vendors include Avena, Autodesk, Dassault Systèmes, GE, and Siemens. Leading CSPs include Amazon, Google, IBM, and Microsoft.

Boost collaboration and communication

Engineers and others struggle to collaborate with staff and external partners due to data-sharing limitations and incompatible technologies. Sometimes, well-intentioned security measures become impediments.

Digital transformation introduces tools that enhance communication and collaboration across teams, departments, and geographical locations.

This interconnectedness reduces bottlenecks, shortens project timelines, and fosters a more agile workplace where engineering teams can collaborate more productively. With the rise of remote work, these tools have become even more critical, allowing businesses to maintain productivity even when their employees are not physically present in their respective offices.

Cloud-based collaboration software includes MindMeister, Miro, Microsoft Teams, Slack and Zoom. Project management software includes Asana, Microsoft Project, and Trello. This software makes it easier for employees to work together in real time, regardless of location.

Bolster agility and flexibility

Many organizations are comfortable with their current processes. That comfort often precludes an agile response when changes in the business environment threaten the business plan.

Digital transformation equips engineers with the agility to respond rapidly to changing market conditions, technological disruptions, and customer needs because the needed data is immediately accessible. In the past, engineers would often have to wait weeks or months to implement responses or launch new products, but digital tools enable them to do so in days or even hours.

For example, cloud computing allows businesses to scale up or down based on demand, enabling them to be more responsive to fluctuations in market needs. Similarly, AI and data analytics enable businesses to pivot quickly based on real-time data insights. This flexibility enhances productivity by ensuring businesses allocate resources optimally and capitalize on opportunities as they arise.

Leading software vendors for data analytics include Google Data Studio, Microsoft Power BI, Minitab, Tableau, TIBCO and TrendMiner. Leading AI software vendors include Anthropic, Google, IBM, Meta Platforms, Microsoft and Open AI.

Empower the workforce and build skills

The workforce, including engineers, frequently feels hemmed in by narrowly defined roles, inadequate digital tools and stifled by a ponderous top-down decision-making culture.

Digital transformation drives productivity by empowering employees with better tools and access to information. With the right digital tools, employees can complete tasks more efficiently, collaborate more easily, experiment and make better decisions.

Digital transformation often opens opportunities for employees to upskill or reskill, enabling them to perform new roles or handle more complex tasks while improving productivity.

Many businesses are investing in training and development programs that teach employees how to use advanced technologies like generative AI, ML, and data analytics tools. This training enhances individual productivity and positions the organization to adapt quickly to technological changes.

Leading immersive training software vendors include AllenComm, EI, ELB Learning, Learning Pool and SweetRush.

Digital transformation is a fundamental shift in how businesses operate and deliver value. It has proven to be a powerful driver of productivity, enabling businesses to streamline processes, automate mundane tasks, make data-driven decisions, enhance customer experience, and boost collaboration and communication. It’s not just a technological upgrade.

Organizations that harness digital transformation’s power will enjoy sustained productivity gains and long-term success. They are better equipped to navigate the complexities of the marketplace, innovate faster, and remain competitive.

The post More digital transformation ideas for raising productivity appeared first on Engineering.com.

Digital transformation has become a vital force in reshaping industries. Adopting and integrating digital technologies into most business areas fundamentally changes how organizations operate and deliver customer value. As engineers worldwide navigate this evolution, one of the compelling benefits of digital transformation is its ability to raise productivity. Through advanced software, automation, data analytics, and enhanced connectivity, digital transformation enables businesses to operate more efficiently, innovate faster, and deliver better outcomes.

Below, we explore the various ways in which digital transformation boosts productivity.

Streamline business processes

Today, engineers spend too much time in low-productivity work such as:

• Hunting for data.
• Seeking access to data.
• Cleaning data to a reasonable level of accuracy and completeness.
• Integrating data using Excel.
• Waiting for others to complete manual work.

Digital transformation delivers significant productivity gains when it includes re-engineering business processes by:

• Simplifying the steps involved.
• Designing steps to minimize the opportunities for errors.
• Improving the availability of relevant digital data to those performing the process.
• Integrating more efficient digital tools into the process.

Digital transformation often triggers the replacement of legacy systems with current software, such as enterprise resource planning (ERP) systems, customer relationship management (CRM) platforms, and industry-specific software-as-a-service (SaaS) solutions. Current software packages and SaaS solutions offer:

• More comprehensive functionality.
• Access to best practice processes.
• Better alignment with various business functions.
• Outsourced software maintenance.

For example, ERP systems consolidate many business processes—such as financial accounting, procurement, production management and supply chain management—into one seamless system, reducing the time spent on redundant tasks and increasing the speed at which businesses can operate and make decisions. Streamlined processes lead to improved staff productivity, fewer delays and higher accuracy.

Leading software to design and automate custom processes include Appian, Microsoft Power Automate, Outsystems, Pegasystems Pega, and Oracle BPM Suite. Leading ERP software vendors include Infor, Microsoft Dynamics 365, Oracle Netsuite, SAP S/4 HANA, and Workday.

Automate repetitive tasks

In many businesses, repetitive tasks consume significant time for engineers. The work is not rewarding and error-prone.

Digital transformation enables the automation of repetitive tasks to drive productivity and data quality. Businesses can automate routine and repetitive tasks, such as data entry, report generation, and many customer service interactions, through the use of artificial intelligence (AI), robotic process automation (RPA), and machine learning (ML).

For example, RPA can streamline back-office operations like finance and HR by processing transactions, aggregating data, and performing multi-step workflows without human intervention. This automation reduces the likelihood of errors and speeds up processes. It also frees up time for employees to focus on more strategic tasks that require critical thinking and creativity that directly contribute to business growth.

Leading software vendors for automating repetitive tasks include Automation Anywhere, Datamatics, SS&C Blue Prism and UiPath.

Support data-driven decision-making

Today, engineers and others make decisions in many businesses based on experience, partial data and hunches. That’s often riskier than it seems.

Digital transformation facilitates collecting and analyzing vast amounts of data, which organizations can leverage to make more informed decisions. Data analytics tools help engineers rapidly analyze performance patterns, customer behaviours, and market trends, assisting businesses to adjust strategies quickly and stay ahead of competitors.

Using data to drive decision-making processes enhances productivity by enabling more precise forecasting, leading to better production management, inventory management, and resource allocation. Data-driven decision-making processes are particularly valuable in industries like retail, manufacturing, and healthcare, where many minor improvements in efficiency can lead to significant cost savings and faster service delivery.

Leading software vendors for data-driven decision-making include Altair, Alteryx, Databricks, IBM Watson Studio, Oracle Analytics, SAP Analytics Cloud and Snowflake. Leading software vendors for data visualization include Google Data Studio, Microsoft Power BI, Minitab, Tableau, TIBCO and TrendMiner.

Enhance the customer experience

Organizations continue to experience difficulties delivering the customer service they aspire to.

A key goal of digital transformation is improving the customer experience in all channels, digital or otherwise. Tools such as AI-powered chatbots, self-service portals, and mobile applications allow businesses to serve customers more efficiently and responsively. Digital transformation enables engineers to interact with customers in real time, providing faster responses to inquiries, better product recommendations, and more personalized experiences.

Improved customer experiences can increase customer satisfaction, loyalty, and repeat business. When it’s possible to assign fewer employees to resolve common customer issues, productivity increases and costs decrease. Dedicating more employees to customer relationship building and innovation drives sales and profitability.

Leading software vendors for customer experience management include Birdeye, HubSpot, Microsoft Dynamics 365 Customer Insights, Podium, and Zendesk. Leading software vendors for call center operation include Five9, Nextiva, Nice CX-One, RingCentral and Talkdesk.

Digital transformation is not just a technological upgrade but a fundamental shift in how businesses operate and deliver value. It has proven to be a powerful driver of productivity, enabling businesses to streamline processes, automate mundane tasks, make data-driven decisions, enhance customer experience, and boost collaboration and communication.

Organizations that embrace digital transformation are better equipped to navigate the complexities of the modern marketplace, innovate faster, and remain competitive. As technology continues to evolve, the businesses that effectively harness the power of digital transformation will enjoy sustained productivity gains and long-term success.

The post How digital transformation raises productivity appeared first on Engineering.com.

That’s easier said than done. The real challenge lies in execution. Starting with a digital transformation roadmap helps turn ambitious goals into tangible digital outcomes that produce business benefits.

A digital transformation roadmap elaborates the high-level strategy into a guide for navigating the complex journey of integrating more digital technology into most business areas. The roadmap bridges the gap between your long-term digital vision and the multiple projects needed to turn the vision into digital reality.

A successful digital transformation roadmap isn’t a plan; it’s a strategic guide for navigating the complex journey of integrating digital technology into every area of your business.

There’s no simple formula for digital transformation or a one-size-fits-all solution. The digital transformation will likely look different even for every company that competes in the same industry. The differences result from various factors, including size, history, in-place technology, degree of centralization, leadership styles and culture.

Digital transformation drivers

The drivers that are causing many organizations to invest in digital transformation include:

• Fewer customers shopping in person and more shopping online.
• More employees working remotely part of most weeks.
• Pressure from competitor actions or to gain competitive advantage.
• Increasing customer expectations for product quality and customer service.
• Increasing product and service complexity requiring more multi-disciplinary collaboration.
• The availability of cheaper and more capable information technology to support digital transformation.
• Improved digital data quality to support digital transformation.

Value of a roadmap

Because digital transformation has a wide-ranging impact on the organization, it’s easy for senior executives, engineers, various stakeholders and project teams to lose track of the vision and strategy. The value of a clear, actionable roadmap includes:

• Ensuring that digital transformation projects are well aligned with business goals.
• Communicating the vision and strategy to maintain organizational commitment to digital transformation.
• Translating sometimes vague digital transformation concepts into multiple clear, actionable objectives that individuals can understand and support.
• Providing an initial insight into the resources the digital transformation projects will require.
• Guiding the projects that will implement various parts of the digital transformation.
• Keeping project teams aligned and focused while understanding their work’s broader digital transformation context.
• Avoiding common pitfalls that impede digital transformation.

Scope of a roadmap

Surprisingly, the scope of digital transformation roadmaps is similar across industries. They typically include many of the following:

• Enhancing the customer experience through technology.
• Leveraging internal and external data with analytics for insightful decision-making.
• Streamlining product design, manufacturing and distribution with process automation.
• Shortening and reducing the complexity of the supply chain through improved digital collaboration with suppliers and dealers.

• Applying available digital technologies to advance the business plan.

Steps in building a roadmap

Engineers typically collaborate with other professionals to build a digital transformation roadmap using the following steps:

1. Define a digital vision for the organization and seek senior executive approval.
2. Develop a digital strategy that captures business process and technology opportunities while recognizing people and facility constraints.
3. Seek senior executive approval for the digital strategy.
4. Confirm the collaboration and people change management strategies.
5. Build a list of digital transformation projects that implement parts of the digital strategy. Ensure every project advances the strategy.
6. Build a high-level plan for every digital transformation project. Ensure no project is planned to run for more than 10 – 12 elapsed months.
7. Describe the required project team skills and experiences. Estimate the approximate effort needed for each role for every digital transformation project.
8. Sequence the digital transformation projects. Prioritize low-complexity, high-value projects earlier. Recognize precedence relationships across projects.
9. Select the appropriate supporting digital technologies and tools. Employ in-place technology where possible.
10. Define the risk management process and the initial risk list with recommended mitigations.
11. Develop a management strategy for project governance.
12. Conduct a short pilot of digitizing a business process using the selected digital technologies and tools to confirm the planned approach.
13. Revise the roadmap based on findings from the pilot.
14. Seek roadmap approval from stakeholders and senior executives.

The digital transformation roadmap implements the strategy that describes the following topics:

• Organization-specific digital transformation drivers.
• The organization’s current state of digitization and automation and how much the digital transformation is expected to advance digital work.
• The approximate amount of business process change, people change, and digital skills upgrading that will be required.

The following categories of digital technologies, tools and data should be considered for the digital transformation projects:

• Advanced data analytics tools.
• Generative AI software.
• Hybrid work management software.
• External data.
• Simulation software.
• SCADA/IIoT systems for operations.
• Software-as-a-Service (SaaS) solutions.
• Custom software development tools.
• Cloud service providers.

Steps in executing a roadmap

Organizations execute the approved digital transformation roadmap using the following steps:

1. Execute the next project in the approved sequence.
2. Review project outcome against project goal and strategy.
3. Refine the roadmap based on project learning and changes in the business environment.
4. Revise project scope definitions based on the previous step.
5. Return to step 1.

Every digital transformation project will include examples of the following deliverables:

• A list of business processes to be improved with rationalization and digital support.
• Additions to the computing infrastructure.
• Additions to the list of datastores.
• Custom software for datastore integration.
• Implementation of apps and SaaS solutions.
• People change management.
• Data quality improvements.

If sufficient resources are available and projects are truly independent, it may be possible to execute two projects concurrently. The benefit will be the earlier start of the benefits of the digital transformation roadmap.

How to recognize a superficial or ineffective roadmap

Engineers can recognize a superficial or ineffective digital transformation roadmap if it contains one or more of the following features:

• Restates the vision and strategy without offering elaboration.
• Discusses technology at length, perhaps to the exclusion of business issues.
• Relies on leading-edge technology.
• Fails to reference the associated business process changes.
• Discusses intangible benefits in detail.
• Omits estimates of tangible benefits.
• Describes the required projects in great detail or not at all.
• Omits a risk and mitigation discussion.
• Contains only cursory references to people change management.
• Proposes a pace of project work that will overwhelm the organization.

A comprehensive roadmap will guide a successful digital transformation and reduce the risk of failure among the projects implementing the strategy.

The post How to build a digital transformation roadmap appeared first on Engineering.com.