Injecting area experience into your AI system | by Dr. Janna Lipenkova | Feb, 2025

Learn how to join the dots between AI expertise and actual life

When beginning their AI initiatives, many corporations are trapped in silos and deal with AI as a purely technical enterprise, sidelining area consultants or involving them too late. They find yourself with generic AI purposes that miss {industry} nuances, produce poor suggestions, and rapidly change into unpopular with customers. Against this, AI programs that deeply perceive industry-specific processes, constraints, and resolution logic have the next advantages:

• Elevated effectivity — The extra area data AI incorporates, the much less handbook effort is required from human consultants.
• Improved adoption — Specialists disengage from AI programs that really feel too generic. AI should communicate their language and align with actual workflows to achieve belief.
• A sustainable aggressive moat — As AI turns into a commodity, embedding proprietary experience is the simplest approach to construct defensible AI programs (cf. this text to study concerning the constructing blocks of AI’s aggressive benefit).

Area consultants might help you join the dots between the technicalities of an AI system and its real-life utilization and worth. Thus, they need to be key stakeholders and co-creators of your AI purposes. This information is the primary a part of my sequence on expertise-driven AI. Following my psychological mannequin of AI programs, it supplies a structured strategy to embedding deep area experience into your AI.

All through the article, we are going to use the use case of provide chain optimisation (SCO) for instance these totally different strategies. Fashionable provide chains are underneath fixed pressure from geopolitical tensions, local weather disruptions, and risky demand shifts, and AI can present the type of dynamic, high-coverage intelligence wanted to anticipate delays, handle dangers, and optimise logistics. Nevertheless, with out area experience, these programs are sometimes disconnected from the realities of life. Let’s see how we will remedy this by integrating area experience throughout the totally different parts of the AI utility.

AI is simply as domain-aware as the info it learns from. Uncooked knowledge isn’t sufficient — it should be curated, refined, and contextualised by consultants who perceive its that means in the true world.

Information understanding: Instructing AI what issues

Whereas knowledge scientists can construct subtle fashions to analyse patterns and distributions, these analyses usually keep at a theoretical, summary stage. Solely area consultants can validate whether or not the info is full, correct, and consultant of real-world circumstances.

In provide chain optimisation, for instance, cargo information might include lacking supply timestamps, inconsistent route particulars, or unexplained fluctuations in transit instances. An information scientist may discard these as noise, however a logistics skilled may have real-world explanations of those inconsistencies. As an illustration, they may be brought on by weather-related delays, seasonal port congestion, or provider reliability points. If these nuances aren’t accounted for, the AI may study an excessively simplified view of provide chain dynamics, leading to deceptive threat assessments and poor suggestions.

Specialists additionally play a essential function in assessing the completeness of information. AI fashions work with what they’ve, assuming that each one key components are already current. It takes human experience and judgment to establish blind spots. For instance, in case your provide chain AI isn’t skilled on customs clearance instances or manufacturing unit shutdown histories, it gained’t be capable of predict disruptions brought on by regulatory points or manufacturing bottlenecks.

✅ Implementation tip: Run joint Exploratory Information Evaluation (EDA) periods with knowledge scientists and area consultants to establish lacking business-critical data, guaranteeing AI fashions work with an entire and significant dataset, not simply statistically clear knowledge.

One widespread pitfall when beginning with AI is integrating an excessive amount of knowledge too quickly, resulting in complexity, congestion of your knowledge pipelines, and blurred or noisy insights. As a substitute, begin with a few high-impact knowledge sources and increase incrementally based mostly on AI efficiency and consumer wants. As an illustration, an SCO system might initially use historic cargo knowledge and provider reliability scores. Over time, area consultants might establish lacking data — resembling port congestion knowledge or real-time climate forecasts — and level engineers to these knowledge sources the place it may be discovered.

✅ Implementation tip: Begin with a minimal, high-value dataset (usually 3–5 knowledge sources), then increase incrementally based mostly on skilled suggestions and real-world AI efficiency.

AI fashions study by detecting patterns in knowledge, however generally, the proper studying indicators aren’t but current in uncooked knowledge. That is the place knowledge annotation is available in — by labelling key attributes, area consultants assist the AI perceive what issues and make higher predictions. Take into account an AI mannequin constructed to foretell provider reliability. The mannequin is skilled on cargo information, which include supply instances, delays, and transit routes. Nevertheless, uncooked supply knowledge alone doesn’t seize the complete image of provider threat — there are not any direct labels indicating whether or not a provider is “excessive threat” or “low threat.”

With out extra specific studying indicators, the AI may make the improper conclusions. It may conclude that each one delays are equally unhealthy, even when some are brought on by predictable seasonal fluctuations. Or it would overlook early warning indicators of provider instability, resembling frequent last-minute order adjustments or inconsistent stock ranges.

Area consultants can enrich the info with extra nuanced labels, resembling provider threat classes, disruption causes, and exception-handling guidelines. By introducing these curated studying indicators, you’ll be able to be certain that AI doesn’t simply memorise previous tendencies however learns significant, decision-ready insights.

You shouldn’t rush your annotation efforts — as a substitute, take into consideration a structured annotation course of that features the next parts:

• Annotation pointers: Set up clear, standardized guidelines for labeling knowledge to make sure consistency. For instance, provider threat classes must be based mostly on outlined thresholds (e.g., supply delays over 5 days + monetary instability = excessive threat).
• A number of skilled assessment: Contain a number of area consultants to scale back bias and guarantee objectivity, notably for subjective classifications like threat ranges or disruption affect.
• Granular labelling: Seize each direct and contextual components, resembling annotating not simply cargo delays but in addition the trigger (customs, climate, provider fault).
• Steady refinement: Repeatedly audit and refine annotations based mostly on AI efficiency — if predictions constantly miss key dangers, consultants ought to regulate labelling methods accordingly.

✅ Implementation tip: Outline an annotation playbook with clear labelling standards, contain at the least two area consultants per essential label for objectivity, and run common annotation assessment cycles to make sure AI is studying from correct, business-relevant insights.

Thus far, our AI fashions study from real-life historic knowledge. Nevertheless, uncommon, high-impact occasions — like manufacturing unit shutdowns, port closures, or regulatory shifts in our provide chain situation — could also be underrepresented. With out publicity to those situations, AI can fail to anticipate main dangers, resulting in overconfidence in provider stability and poor contingency planning. Artificial knowledge solves this by creating extra datapoints for uncommon occasions, however skilled oversight is essential to make sure that it displays believable dangers relatively than unrealistic patterns.

Let’s say we wish to predict provider reliability in our provide chain system. The historic knowledge might have few recorded provider failures — however that’s not as a result of failures don’t occur. Relatively, many corporations proactively mitigate dangers earlier than they escalate. With out artificial examples, AI may deduce that provider defaults are extraordinarily uncommon, resulting in misguided threat assessments.

Specialists might help generate artificial failure situations based mostly on:

• Historic patterns — Simulating provider collapses triggered by financial downturns, regulatory shifts, or geopolitical tensions.
• Hidden threat indicators — Coaching AI on unrecorded early warning indicators, like monetary instability or management adjustments.
• Counterfactuals — Creating “what-if” occasions, resembling a semiconductor provider all of the sudden halting manufacturing or a chronic port strike.

✅ Actionable step: Work with area consultants to outline high-impact however low-frequency occasions and situations, which may be in focus once you generate artificial knowledge.

Information makes area experience shine. An AI initiative that depends on clear, related, and enriched area knowledge could have an apparent aggressive benefit over one which takes the “quick-and-dirty” shortcut to knowledge. Nevertheless, take into account that working with knowledge may be tedious, and consultants have to see the end result of their efforts — whether or not it’s bettering AI-driven threat assessments, optimising provide chain resilience, or enabling smarter decision-making. The bottom line is to make knowledge collaboration intuitive, purpose-driven, and immediately tied to enterprise outcomes, so consultants stay engaged and motivated.

As soon as AI has entry to high-quality knowledge, the following problem is guaranteeing it generates helpful and correct outputs. Area experience is required to:

1. Outline clear AI goals aligned with enterprise priorities
2. Guarantee AI accurately interprets industry-specific knowledge
3. Repeatedly validate AI’s outputs and proposals

Let’s have a look at some widespread AI approaches and see how they will profit from an additional shot of area data.

Coaching predictive fashions from scratch

For structured issues like provide chain forecasting, predictive fashions resembling classification and regression might help anticipate delays and counsel optimisations. Nevertheless, to verify these fashions are aligned with enterprise targets, knowledge scientists and data engineers have to work collectively. For instance, an AI mannequin may attempt to minimise cargo delays in any respect prices, however a provide chain skilled is aware of that fast-tracking each cargo by means of air freight is financially unsustainable. They’ll formulate further constraints on the mannequin, making it prioritise essential shipments whereas balancing value, threat, and lead instances.

✅ Implementation tip: Outline clear goals and constraints with area consultants earlier than coaching AI fashions, guaranteeing alignment with actual enterprise priorities.

For an in depth overview of predictive AI strategies, please confer with Chapter 4 of my ebook The Artwork of AI Product Administration.

Navigating the LLM triad

Whereas predictive fashions skilled from scratch can excel at very particular duties, they’re additionally inflexible and can “refuse” to carry out every other activity. GenAI fashions are extra open-minded and can be utilized for extremely numerous requests. For instance, an LLM-based conversational widget in an SCO system can permit customers to work together with real-time insights utilizing pure language. As a substitute of sifting by means of rigid dashboards, customers can ask, “Which suppliers are liable to delays?” or “What different routes can be found?” The AI pulls from historic knowledge, stay logistics feeds, and exterior threat components to supply actionable solutions, counsel mitigations, and even automate workflows like rerouting shipments.

However how can you make sure that an enormous, out-of-the-box mannequin like ChatGPT or Llama understands the nuances of your area? Let’s stroll by means of the LLM triad — a development of strategies to include area data into your LLM system.

As you progress from left to proper, you’ll be able to ingrain extra area data into the LLM — nevertheless, every stage additionally provides new technical challenges (in case you are serious about a scientific deep-dive into the LLM triad, please take a look at chapters 5–8 of my ebook The Artwork of AI Product Administration). Right here, let’s give attention to how area consultants can soar in at every of the levels:

1. Prompting out-of-the-box LLMs may seem to be a generic strategy, however with the proper instinct and ability, area consultants can fine-tune prompts to extract the additional little bit of area data out of the LLM. Personally, I believe it is a large a part of the fascination round prompting — it places probably the most highly effective AI fashions immediately into the palms of area consultants with none technical experience. Some key prompting strategies embrace:

• Few-shot prompting: Incorporate examples to information the mannequin’s responses. As a substitute of simply asking “What are different transport routes?”, a well-crafted immediate contains pattern situations, resembling “Instance of previous situation: A earlier delay on the Port of Shenzhen was mitigated by rerouting by means of Ho Chi Minh Metropolis, lowering transit time by 3 days.”
• Chain-of-thought prompting: Encourage step-by-step reasoning for advanced logistics queries. As a substitute of “Why is my cargo delayed?”, a structured immediate may be “Analyse historic supply knowledge, climate experiences, and customs processing instances to find out why cargo #12345 is delayed.”
• Offering additional background data: Connect exterior paperwork to enhance domain-specific responses. For instance, prompts may reference real-time port congestion experiences, provider contracts, or threat assessments to generate data-backed suggestions. Most LLM interfaces already let you conveniently connect further information to your immediate.

2. RAG (Retrieval-Augmented Era): Whereas prompting helps information AI, it nonetheless depends on pre-trained data, which can be outdated or incomplete. RAG permits AI to retrieve real-time, company-specific knowledge, guaranteeing that its responses are grounded in present logistics experiences, provider efficiency information, and threat assessments. For instance, as a substitute of producing generic provider threat analyses, a RAG-powered AI system would pull real-time cargo knowledge, provider credit score rankings, and port congestion experiences earlier than making suggestions. Area consultants might help choose and construction these knowledge sources and are additionally wanted in terms of testing and evaluating RAG programs.

✅ Implementation tip: Work with area consultants to curate and construction data sources — guaranteeing AI retrieves and applies solely probably the most related and high-quality enterprise data.

3. Fantastic-tuning: Whereas prompting and RAG inject area data on-the-fly, they don’t inherently embed provide domain-specific workflows, terminology, or resolution logic into your LLM. Fantastic-tuning adapts the LLM to suppose like a logistics skilled. Area consultants can information this course of by creating high-quality coaching knowledge, guaranteeing AI learns from actual provider assessments, threat evaluations, and procurement choices. They’ll refine {industry} terminology to stop misinterpretations (e.g., AI distinguishing between “buffer inventory” and “security inventory”). Additionally they align AI’s reasoning with enterprise logic, guaranteeing it considers value, threat, and compliance — not simply effectivity. Lastly, they consider fine-tuned fashions, testing AI in opposition to real-world choices to catch biases or blind spots.

✅ Implementation tip: In LLM fine-tuning, knowledge is the essential success issue. High quality goes over amount, and fine-tuning on a small, high-quality dataset may give you glorious outcomes. Thus, give your consultants sufficient time to determine the proper construction and content material of the fine-tuning knowledge and plan for loads of end-to-end iterations of your fine-tuning course of.

Encoding skilled data with neuro-symbolic AI

Each machine studying algorithm will get it improper occasionally. To mitigate errors, it helps to set the “exhausting information” of your area in stone, making your AI system extra dependable and controllable. This mix of machine studying and deterministic guidelines is known as neuro-symbolic AI.

For instance, an specific data graph can encode provider relationships, regulatory constraints, transportation networks, and threat dependencies in a structured, interconnected format.

As a substitute of relying purely on statistical correlations, an AI system enriched with data graphs can:

• Validate predictions in opposition to domain-specific guidelines (e.g., guaranteeing that AI-generated provider suggestions adjust to regulatory necessities).
• Infer lacking data (e.g., if a provider has no historic delays however shares dependencies with high-risk suppliers, AI can assess its potential threat).
• Enhance explainability by permitting AI choices to be traced again to logical, rule-based reasoning relatively than black-box statistical outputs.

How are you going to determine which data must be encoded with guidelines (symbolic AI), and which must be discovered dynamically from the info (neural AI)? Area consultants might help youpick these bits of information the place hard-coding makes probably the most sense:

• Data that’s comparatively steady over time
• Data that’s exhausting to deduce from the info, for instance as a result of it’s not well-represented
• Data that’s essential for high-impact choices in your area, so you’ll be able to’t afford to get it improper

Normally, this information will likely be saved in separate parts of your AI system, like resolution bushes, data graphs, and ontologies. There are additionally some strategies to combine it immediately into LLMs and different statistical fashions, resembling Lamini’s reminiscence fine-tuning.

Compound AI and workflow engineering

Producing insights and turning them into actions is a multi-step course of. Specialists might help you mannequin workflows and decision-making pipelines, guaranteeing that the method adopted by your AI system aligns with their duties. For instance, the next pipeline exhibits how the AI parts we thought of to date may be mixed right into a workflow for the mitigation of cargo dangers:

Specialists are additionally wanted to calibrate the “labor distribution” between people in AI. For instance, when modelling resolution logic, they will set thresholds for automation, deciding when AI can set off workflows versus when human approval is required.

✅ Implementation tip: Contain your area consultants in mapping your processes to AI fashions and property, figuring out gaps vs. steps that may already be automated.

Particularly in B2B environments, the place staff are deeply embedded of their every day workflows, the consumer expertise should be seamlessly built-in with present processes and activity constructions to make sure effectivity and adoption. For instance, an AI-powered provide chain software should align with how logistics professionals suppose, work, and make choices. Within the improvement section, area consultants are the closest “friends” to your actual customers, and selecting their brains is among the quickest methods to bridge the hole between AI capabilities and real-world usability.

✅ Implementation tip: Contain area consultants early in UX design to make sure AI interfaces are intuitive, related, and tailor-made to actual decision-making workflows.

Guaranteeing transparency and belief in AI choices

AI thinks otherwise from people, which makes us people skeptical. Usually, that’s a great factor because it helps us keep alert to potential errors. However mistrust can be one of many largest limitations to AI adoption. When customers don’t perceive why a system makes a specific suggestion, they’re much less prone to work with it. Area consultants can outline how AI ought to clarify itself — guaranteeing customers have visibility into confidence scores, resolution logic, and key influencing components.

For instance, if an SCO system recommends rerouting a cargo, it could be irresponsible on the a part of a logistics planner to simply settle for it. She must see the “why” behind the advice — is it attributable to provider threat, port congestion, or gas value spikes? The UX ought to present a breakdown of the choice, backed by further data like historic knowledge, threat components, and a cost-benefit evaluation.

⚠️ Mitigate overreliance on AI: Extreme dependence of your customers on AI can introduce bias, errors, and unexpected failures. Specialists ought to discover methods to calibrate AI-driven insights vs. human experience, moral oversight, and strategic safeguards to make sure resilience, adaptability, and belief in decision-making.

✅ Implementation tip: Work with area consultants to outline key explainability options — resembling confidence scores, knowledge sources, and affect summaries — so customers can rapidly assess AI-driven suggestions.

Simplifying AI interactions with out shedding depth

AI instruments ought to make advanced choices simpler, not tougher. If customers want deep technical data to extract insights from AI, the system has failed from a UX perspective. Area consultants might help strike a steadiness between simplicity and depth, guaranteeing the interface supplies actionable, context-aware suggestions whereas permitting deeper evaluation when wanted.

As an illustration, as a substitute of forcing customers to manually sift by means of knowledge tables, AI may present pre-configured experiences based mostly on widespread logistics challenges. Nevertheless, skilled customers also needs to have on-demand entry to uncooked knowledge and superior settings when obligatory. The bottom line is to design AI interactions which can be environment friendly for on a regular basis use however versatile for deep evaluation when required.

✅ Implementation tip: Use area skilled suggestions to outline default views, precedence alerts, and user-configurable settings, guaranteeing AI interfaces present each effectivity for routine duties and depth for deeper analysis and strategic choices.

Steady UX testing and iteration with consultants

AI UX isn’t a one-and-done course of — it must evolve with real-world consumer suggestions. Area consultants play a key function in UX testing, refinement, and iteration, guaranteeing that AI-driven workflows keep aligned with enterprise wants and consumer expectations.

For instance, your preliminary interface might floor too many low-priority alerts, resulting in alert fatigue the place customers begin ignoring AI suggestions. Provide chain consultants can establish which alerts are most beneficial, permitting UX designers to prioritize high-impact insights whereas lowering noise.

✅ Implementation tip: Conduct think-aloud periods and have area consultants verbalize their thought course of when interacting together with your AI interface. This helps AI groups uncover hidden assumptions and refine AI based mostly on how consultants truly suppose and make choices.

Vertical AI programs should combine area data at each stage, and consultants ought to change into key stakeholders in your AI improvement:

• They refine knowledge choice, annotation, and artificial knowledge.
• They information AI studying by means of prompting, RAG, and fine-tuning.
• They assist the design of seamless consumer experiences that combine with every day workflows in a clear and reliable manner.

An AI system that “will get” the area of your customers is not going to solely be helpful and adopted within the short- and middle-term, but in addition contribute to the aggressive benefit of your enterprise.

Now that you’ve got discovered a bunch of strategies to include domain-specific data, you may be questioning learn how to strategy this in your organizational context. Keep tuned for my subsequent article, the place we are going to contemplate the sensible challenges and methods for implementing an expertise-driven AI technique!

Observe: Except famous in any other case, all photos are the writer’s.