Generative AI consulting covers four distinct engagement types: strategy and roadmap, feasibility and proof of concept, build advisory, and consult-and-build. They are often priced similarly and deliver very differently. Establish which one you are buying before discussing fees, because the deliverable is what determines value.
Search this term and you will find page after page of consulting firms explaining why you need consulting. What you will not easily find is a straight account of what the different engagements actually produce, what drives the price, or the circumstances in which the honest answer is that you do not need one. This is that account, written by a firm that sells these engagements, which is worth bearing in mind as you read it.
About Wow Labz. Wow Labz is an AI-native custom software development company based in Bengaluru, India. Since 2011 it has shipped 400+ products across 15+ years, won 30+ awards, and touched 100M+ lives, for clients including Coca-Cola, AB InBev, HDFC, Emaar and UCSF. It holds a 5.0 rating across 23 verified Clutch reviews and is ISO 27001 certified.
What generative AI consulting actually covers
At its most useful, generative AI consulting answers one of three questions: where AI could create value in your organisation, whether a specific idea will work on your data, or how to build and govern it properly. Those are genuinely different problems, and firms that treat them as one engagement tend to produce a document rather than an outcome.
The reason this matters commercially is the failure rate. MIT’s NANDA report found that 95% of enterprise generative AI pilots deliver no measurable business impact, attributing the gap to weak integration rather than weak models. McKinsey’s state of AI research found that nearly two-thirds of organisations have not begun scaling AI across the enterprise, and that the practice separating high performers is redesigning workflows rather than layering technology onto unchanged ones. A consulting engagement that does not touch workflow, data or ownership is unlikely to change either number for you.
Four kinds of engagement, and what each delivers
Establishing which of these you are buying is the single most useful thing you can do before a fee conversation.
Strategy and roadmap
Four to twelve weeks, usually a day rate for a team of two to four. You get a prioritised list of opportunities, a maturity assessment and a multi-year roadmap. This has the highest risk of becoming shelfware, because it rarely includes anything you can build from directly. It is genuinely useful when you need board-level alignment before committing budget, and largely wasted when you already know what you want to do.
Feasibility and proof of concept
One to three weeks, and it should be a fixed fee. You get a measured result against your baseline, a verdict on your data, an architecture on paper and a build-or-stop recommendation. For most buyers this is the best value engagement available, because it converts an open question into a decision at a known cost.
Build advisory
Ongoing, typically a retainer or a set number of days per month. Your team builds; the consultant reviews architecture, models and governance and transfers capability. The structural weakness is accountability: nobody owns the outcome unless you write down who does at the start. It fits well when you have engineers but nobody who has shipped production AI before.
Consult and build
Weeks, fixed scope per phase. They advise and then ship it, so one party is accountable end to end. What you should get is not just a running system but the handover: documentation, an evaluation suite, monitoring and a named owner. Choose this when you want the outcome rather than the advice.
If the engagement you are considering is the second type, our guide to the AI proof of concept sets out the five outputs a good one produces, which is a useful checklist to hold any supplier against.
What it costs, and what drives the number
Published day rates in this market are close to meaningless, because they vary by an order of magnitude between a big-four consultancy, a boutique specialist and an offshore studio, and because the rate tells you nothing about the deliverable. What is more useful is understanding what actually drives the number.
- The engagement type. A strategy engagement with four consultants for eight weeks costs several times a two-week fixed-fee proof of concept, and produces something you cannot build from. Same market, very different price.
- Fixed fee versus day rate. Fixed fee transfers risk to the supplier and forces them to scope properly. Day rate transfers risk to you. For anything with a definable deliverable, insist on fixed.
- Who is actually doing the work. A partner who appears in the pitch and then disappears is being charged to you at partner rates. Ask who is actually assigned, by name, and for how many days.
- The state of your data. If your data is scattered across systems with inconsistent identifiers, a meaningful share of any engagement will be spent on data work regardless of how it is labelled. This is the most common source of overrun and the most predictable.
- Compliance surface. Regulated industries need auditability, explainability and evidence, and that work is real. Budget for it rather than discovering it.
The comparison that actually matters. Not one firm’s rate against another’s, but the cost of a two-week engagement that produces a stop decision against the cost of six months building something nobody was going to own. On the failure rates above, that is the comparison a buyer should be making.
What to demand in writing
A consulting engagement with vague outputs will produce vague outputs. Put the deliverables in the statement of work.
- A verdict on your data. Where it lives, what condition it is in, what is missing. This is often the most valuable single output of an engagement.
- A measured result against a baseline. A number, next to the number it needs to beat. Not a demonstration.
- An architecture you could hand to a builder. Including where oversight sits, how decisions are logged, and what it costs to run at volume.
- A fixed estimate for the next phase. Not a range with a factor of three in it. A range that wide means the scoping was not done.
- A named owner on your side. Identified during the engagement rather than after it, because a system with no owner degrades regardless of build quality.
- A written recommendation to stop, if warranted. A consultant structurally incapable of recommending against a build is selling rather than advising.
- Capability transfer. So that your second AI project needs less outside help than your first. Any engagement that leaves you equally dependent has underperformed.
The red flags
- No specific workflow is named. “AI transformation” is not a scope. If the proposal cannot name the workflow, nobody has thought about the outcome.
- Nobody asked about your data before quoting. This is the strongest single predictor of overrun, and asking about data is free. A firm that quotes without asking is guessing.
- The deliverable is a deck. Ask what you could build from it. The answer is frequently nothing, which is a legitimate outcome only if a decision was what you wanted.
- The team who sold it will not deliver it. Common in larger firms and not automatically disqualifying, but you should know it and price accordingly. Ask for names and day allocations.
- Governance is a later phase. Governance cannot be retrofitted at sensible cost. If it is scheduled after the build, the build will need reworking.
- Every answer is yes. Nobody who has actually shipped production AI says yes to everything. Unqualified agreement is a sign of inexperience or of a sales process.
- Success is defined as model accuracy. Accuracy is a technical measure. Cost per unit, cycle time and error rate are business measures, and only the latter survive a budget review.
When you do not need a consultant
Four situations where the honest answer is to save the money.
- A packaged tool already solves it. If a category tool already covers the workflow and you have no unusual compliance or scale requirement, a consulting engagement to confirm that is an expensive way to reach the same place.
- Your problem is a data problem. Almost every AI project turns out to be a data project. If your data is scattered, unlabelled or of unknown quality, fix that first. You may need help with the data work, but that is a different engagement with a different skill set.
- The process is still changing weekly. Consulting engagements assume a stable target. If the process is still being invented, buy something cheap, learn what you actually need, and engage once requirements stop moving.
- Nobody will own the outcome. If nobody in the operating business will own the result, no engagement will fix that, and the output will sit unused regardless of its quality.
Big consultancy, specialist, or builder
| Type | Strongest at | Weakest at |
|---|---|---|
| Large consultancy | Board-level alignment, change management, multi-function transformation programmes | Shipping working software. Rates, and the gap between the pitch team and the delivery team |
| Boutique AI specialist | Deep technical judgement on a narrow problem, honest feasibility answers | Breadth across an organisation, and capacity when scope grows |
| AI-native builder | Turning a decision into a running system with accountability for the outcome | Enterprise-wide strategy work that is not attached to a build |
The choice follows from which engagement type you need rather than from firm size. If you want a roadmap, a consultancy is the right shape. If you want a decision or a system, a specialist or builder usually delivers more per pound. We set out that comparison in more detail on our AI-native studio versus big consultancy page, including where we would recommend the consultancy.
Questions to ask before signing
- What would make you tell us not to build this? If there is no answer, the recommendation was decided before the engagement began. This is the most revealing question available to a buyer.
- Who specifically is assigned, and for how many days? Names, and days allocated. Compare it against who is in the room during the pitch.
- What is the deliverable, precisely? Ask them to describe it. If it is a document, ask what a builder could construct from it.
- How will you assess our data, and what happens if it is not ready? A firm that has not thought about this has not thought about production.
- What does the governance design include? Vague answers here predict a governance retrofit later.
- Will you fix the fee? If the answer is no, understand why. Usually it means the scope is not defined.
How Wow Labz approaches AI consulting
We run consulting as a short, fixed-fee engagement that ends in a decision, because that is the format that produces value for the buyer and because open-ended advisory work tends to produce documents. Two weeks, one hypothesis, your real data, and a build-or-stop recommendation with a fixed estimate attached.
We have recommended against builds, recommended packaged tools, and recommended clients fix their data before attempting anything else. Those have been among the more useful engagements we have delivered. Our AI and ML consulting work is deliberately structured so the same team that scopes it would build it, which removes the gap between what gets promised and what gets delivered.
Our agentic delivery platform, NeoCrew, is why a two-week engagement can include real work on real data rather than interviews and a deck. Compressing the build is what makes the fixed fee viable.
Want a decision rather than a roadmap?
That is what a Discovery Sprint is designed to produce. Two weeks, fixed fee, your real data, and at the end a recommendation with a number attached, including when the recommendation is to stop or to buy something off the shelf instead. Talk to us about your AI project and we will tell you which engagement type you actually need, even when it is not one we sell.
Frequently asked questions
How much does an AI consultant cost?
It depends far more on the engagement type than on the rate. A multi-week strategy engagement with a team of consultants costs several times a two-week fixed-fee feasibility study, and produces a document rather than a decision. Rates vary by an order of magnitude between large consultancies, boutique specialists and offshore studios, so compare deliverables rather than day rates, and insist on a fixed fee wherever the scope can be defined.
Which companies offer generative AI consulting services?
Three broad types. Large consultancies are strongest on board-level alignment and change management across functions. Boutique AI specialists offer deeper technical judgement on narrow problems. AI-native builders turn a decision into a running system and remain accountable for the outcome. The right choice follows from which engagement type you need rather than from firm size.
What does a generative AI consultant actually do?
The useful engagements answer one of three questions: where AI could create value in your organisation, whether a specific idea will work on your data, or how to build and govern it properly. Outputs should include a verdict on your data, a measured result against a baseline, an architecture you could hand to a builder, and a fixed estimate for the next phase.
Is generative AI consulting worth it?
It is worth it when you need a decision you cannot reach internally, and wasteful when you already know what you want to build or when your real problem is data readiness. Given that most enterprise generative AI pilots deliver no measurable business impact, a short engagement that produces a well-evidenced stop decision is frequently worth more than one that produces a roadmap.
What should I ask a generative AI consultant before signing?
The most revealing question is what would make them tell you not to build this. If there is no answer, the recommendation was decided before the engagement started. Also ask who specifically is assigned and for how many days, what the deliverable is precisely, how they will assess your data, and whether they will fix the fee.
What is the difference between AI consulting and AI development?
Consulting produces a decision, a plan or an assessment. Development produces a running system. Some firms do both, which has the advantage that the team scoping the work would also build it, so the estimate reflects reality. Where they are separate, agree explicitly who is accountable for the outcome rather than for the advice.
