Big consultancy or boutique AI agency? How to decide

In AI consulting, a large consultancy is the sensible choice when the programme touches several countries, business units and regulators at once, and the board wants an independent, institutional signature on it. A boutique AI agency comes out ahead when you need a measurable result on a single process within weeks, and the team that writes the strategy also has to build the system. An in-house team is the right answer once AI has become a permanent capability that needs constant upkeep; for most companies the healthiest route is a mix — start the work with outside help and keep the knowledge inside.

I put the short answer first because the long one has to begin with a disclosure: whoever wrote this is not neutral. INDOLES sits on the boutique side of the line — a small team doing strategy and software under one roof, with our package prices published openly. That is why what follows sets out both the situations in which a large consultancy genuinely is the better choice and the real risks of the boutique side, which we carry too. I have not named any firm: large consultancies are discussed here as a category rather than compared company by company, and the firms within the category differ from one another as well.

This article is the "who should we work with" part of the AI consulting decision set. How to test whichever side you pick in the first meeting is covered in 12 questions to ask an AI consultant; how the first ninety days run once the decision is made is in the 90-day pilot framework; and the scope of our AI consulting service is written out on the service page.

What is the difference between AI consulting and traditional management consulting?

Traditional management consulting clarifies a decision: which market to enter, which investment to make, how to organise — it recommends through analysis and scenarios, and its output is usually a report and a roadmap. AI consulting tests whether AI produces a measurable return in a specific process and, if it does, builds the system; its output is less a report than a measured pilot and a working system.

The two kinds of work part ways along four axes:

• Unit: management consulting works on a decision, AI consulting on a process. The first asks "what should we do?"; the second asks "does AI work in this process?"
• Output: management consulting delivers a document; AI consulting delivers a system that has run on real data, with its result measured side by side against the current method.
• Team: the first leans on analysts and sector specialists; the second needs, alongside process knowledge, the engineering capacity to do the data preparation, the integration and the interface.
• Measurement: management consulting estimates the expected impact of its scenarios; AI consulting measures the baseline before starting and repeats the same measurement after the pilot.

Which one to choose follows from the same axes. If your question is about the company's direction — a new line of business, an acquisition, the organisation structure — what you need is management consulting, and AI may be one of the inputs to that decision. If your question is whether a specific piece of work can be done faster, with fewer errors or at lower cost, what you need is AI consulting. If both questions are on the table at once, order matters: the direction is decided first, and the pilot opens inside that direction.

The line is not as sharp as it used to be. Large consultancies now offer AI consulting too, while some boutique teams build tools without touching strategy at all. So the decision should rest not on the "consulting or AI" label but on which question the team across the table is set up to answer.

What is the difference between a large consultancy and a boutique AI agency?

The difference lies not in quality but in scale and structure. A large consultancy brings a deep bench of specialists, adjacent expertise such as risk, compliance and often legal, and a way of working that is used to corporate governance; a boutique agency brings a small, senior team, a short chain of decisions and, more often than not, a team that does the implementation itself.

"Boutique" doesn't describe a single thing either. The category holds two different kinds of provider: AI agencies that build a chatbot or a content automation on top of an off-the-shelf model, and small consulting and engineering teams that measure the process and build the system around that measure. We set out the difference in the first section of the 12 questions to ask an AI consultant; here, by boutique I mean the second kind, because that is the one answering the same question as a large consultancy.

The decision becomes clear across eight criteria. Each line sets the three options side by side; look at which one describes your situation.

1. Company scale: a large consultancy is strong in multi-country, multi-unit structures. A boutique agency is strong in work at the level of one business unit or one process. An in-house team makes sense at the scale where AI produces work continuously across several units.
2. Regulation and corporate governance: a large firm can bring risk, compliance and legal expertise under the same roof and works through a process an audit committee recognises. A boutique agency usually handles this alongside your own legal adviser. An in-house team ties compliance responsibility to the company's own compliance function.
3. Speed: a large firm's strength is parallel capacity — it can run teams in several units at once. A boutique agency's strength is how fast it reaches the first measurable result on one process, because its decision chain is short. An in-house team's speed is capped by hiring; no work starts until the team exists.
4. Budget and pricing: a large firm's proposal is usually prepared after a scoping conversation and tied to team size and duration. A boutique agency's proposal is tied to scope as well; some publish the prices of fixed-scope packages openly. An in-house team's cost is worked out with salaries plus time to hire, tools and the periods when the team is waiting for work.
5. Who does the implementation: at a large firm, implementation may sit with the firm's own technology unit, with a technology partner or with your team; the proposal should say which. At a boutique agency that keeps strategy and software in one team, implementation stays with the team that wrote the recommendation. With an in-house team, implementation is already inside — but so are process selection and strategy, which then have to be done in-house too.
6. Team continuity: at a large firm, the question to ask is whether the senior people presenting the proposal are the people who will do the work. At a boutique agency, the question is whether the work hangs on one or two people, and what happens when they are unavailable. With an in-house team, the risk is a key person leaving.
7. Independence: with both outside options, ask whether there is a partnership with a software, cloud or model provider; a partnership is not bad in itself, but it can shape the recommendation and should be in writing. An in-house team is independent, but may lean towards the technology it already knows.
8. Knowledge transfer: at a large firm and a boutique agency alike, handover only happens if it is written into the proposal as its own line: source code, rule sets, data, accounts, and two people able to run the system. With an in-house team the knowledge is already inside; the risk is that it never gets written down.

When is a large consultancy the better choice?

A large consultancy is the better choice when the AI decision reaches beyond a single process and the scope is broader than a boutique team can carry. Five situations are typical of this, and in them choosing a boutique team does not save budget — it postpones the risk.

1. The programme spans several countries or units: in a transformation that changes the roles of hundreds of people, change management, training and internal communication are as big as the work itself, and they need parallel team capacity.
2. The sector is heavily regulated: in banking, insurance, healthcare or the public sector, model risk, the audit trail and the legal assessment may need to come from under one roof.
3. The board wants an independent, recognised signature: for some decisions, who carried out the assessment is as much of an assurance to the board and investors as the assessment itself.
4. AI is part of a bigger decision: in a merger, a full ERP replacement or a rebuilt organisation structure, AI is a sub-heading; the direction is decided first.
5. The company is the Türkiye unit of a global group: the method, approved supplier list and sign-off process set by headquarters can narrow the choice from the start.

In the first of these situations, we are not the right address either: the out-of-scope list of our digital transformation service names internal change management and HR consulting. In the second we can build the pilot, but the legal and compliance opinion has to come from your legal adviser rather than from us. I set out the limits of our scope separately in the INDOLES section below; a provider that writes down its limits is easier to assess than one that says it does everything.

When is a boutique AI agency the better choice?

A boutique AI agency is the better choice when the problem can be pinned to a single process, the result has to be measured within weeks, and the team writing the recommendation also has to build the system. Under those conditions a large structure's capacity goes unused, while the hand-off points in between generate cost.

1. The problem sits in one process: repeated, recorded work such as preparing quotes, entering orders, searching technical documents or classifying incoming requests.
2. No hand-off should come between strategy and implementation: if the team that writes the recommendation also does the data preparation, the integration and the interface, the risk of the report gathering dust disappears.
3. The budget needs to be tied to a fixed scope: the scope, duration and exclusions of a single-scenario pilot can be written down before signing.
4. The decision chain is short: in work the founder or managing director owns directly, the person making the weekly decisions and the team doing the work can sit at the same table.
5. Working directly with senior people matters: in a boutique team, the person who comes to the meeting is usually the person who does the work.

Let me show what staying in one team means with an example from our side. At Meccanotecnica Umbra Türkiye, the same team built the product catalogue, the AI technical advisor where an engineer describes their plant and finds the right equipment, the quote portal and the CRM connection; the project ran 22 weeks, and the pieces went live in the same release rather than one by one. Quote requests rose tenfold and the time between request and response fell by ninety percent. Size did not produce that result; a team that knew the catalogue and the advisor had to work together, and built both, did.

When does it make sense to build an in-house AI team?

An in-house team makes sense once AI stops being a project in the company and becomes a capability that needs constant upkeep: when several systems are running live, when models and rule sets are updated regularly, and when process knowledge is itself what sets the company apart from its competitors. A team hired before that point is, more often than not, a team that does not yet know what it is there to do.

In Türkiye the barrier is often expertise before budget. In the Artificial Intelligence Statistics bulletin that TÜİK, the Turkish Statistical Institute, published on 1 October 2025, the reason cited most often by enterprises considering AI but not yet using it was a lack of relevant expertise in the business. An in-house team is the permanent way to close that gap, but not the fastest: hiring can take months, and a data specialist hired on their own cannot carry a pilot without a process owner and an engineer to write the integration.

Which blocks of a pilot your own team can run is clear as well. Process selection, ownership and the baseline need process knowledge rather than AI expertise; outside support is usually needed in the build and field test — model selection, integration and prototype development. We set out that split block by block in the 90-day pilot framework.

How do you set up a mixed model?

A mixed model is an arrangement in which an in-house team takes over work an outside team started, or two outside parties take on different layers — and it runs on one rule: the owner of the result sits inside the company. The process owner comes from the company and the outside parties produce numbers for that person; if this rule is not written down, the mixed model turns into a gap in which neither side owns the result.

Three mixed arrangements stand out:

• A boutique team starts, the in-house team takes over: the outside team builds the pilot; during it at least two people on your team learn the system by working on it alongside, and the source code, rule sets, data and accounts stay with the company when the pilot ends. For most companies this is the healthiest mix.
• A large firm sets the direction, a boutique team implements: the corporate roadmap comes from a large consultancy, and individual pilots are built inside that roadmap by a boutique team. In this arrangement both sides should receive, in the same document, the measure by which the roadmap will be tracked.
• The in-house team manages, outside teams build specific pieces: the company's own AI lead runs the portfolio, and individual pilots that exceed its capacity or expertise go to outside teams.

What is the real risk of each option?

Each option's risk is the shadow of its greatest strength. The large firm's risk comes from its scale, the boutique team's from its smallness, the in-house team's from its isolation, and the mixed model's from nobody owning it.

The risk with a large consultancy

The senior team presenting the proposal and the team doing the work may be different people; if their names are not in the contract, who sits at the table in month one is anyone's guess. The output can stay a report: even if the report is right, without people in the company to implement it the decision waits on the shelf. For work covering a single process, scope, duration and cost can grow out of proportion. If the firm has a partnership with a technology provider, it is your job to ask whether the recommendation leans towards that provider.

The risk with a boutique AI agency

The work may hang on one or two senior people; when they fall ill, leave or get split across three clients at once, the schedule slips. Capacity has a ceiling: running pilots in five units at once is not work a boutique team can carry. Adjacent expertise such as legal, compliance and change management is often missing from the team. The small company's own continuity is a risk too, which is why the system and the knowledge staying with you when the project ends is not up for negotiation on the boutique side — it is a precondition. Every one of these risks applies to us as well: working under one roof means depending on a single point.

The risk with an in-house team

Hiring takes time, and the first pilot waits until the team is in place. A one-person "AI team" is left on its own without a process owner and an engineer to write the integration, and ends up producing trial projects. The model layer is the fastest-changing part of this field; a team with no outside contact can see its knowledge age quickly. The knowledge stays inside, but if it is never written down it walks out of the door with the key person.

The risk with a mixed model

Neither side may own the result: the outside team defends the build, the in-house team defends the running, and the measurement in between becomes nobody's job. If the handover is not written down it never happens; it belongs in the contract as a date in the calendar, two names on the team and four items to be delivered.

Which questions should you ask before signing?

The questions to ask before signing are the ones that bring the differences between the options to the surface, and they should be put to both outside parties in the same way. The eight questions below work for a large consultancy and a boutique team alike; set the answers side by side and the decision usually becomes clear on its own.

1. Are the people presenting the proposal the people who will do the work, and are their names in the contract?
2. Is the output a document or a working system, and who does the implementation?
3. Which baseline will be measured before the project, and is that measurement a line in the proposal?
4. What is the price tied to — a fixed scope, or team and duration — and are the exclusions in writing?
5. Do you have a partnership or a commission arrangement with any software, cloud or model provider?
6. Who ends up with the source code, rule sets, data and accounts when the project closes?
7. Who on our team will be able to run the system without you, and by what date?
8. Whose responsibility is the legal and compliance assessment?

The first and fifth questions tell you most on the large-firm side, the sixth and seventh on the boutique side; the eighth often tells you on its own which side you are actually looking for. Questions about the AI project itself — data preparation, what happens when the model is wrong, provider lock-in — sit separately in the 12 questions article.

How does business development consulting differ from management consulting and from an advertising agency?

An advertising agency moves demand: it produces campaigns, media and creative. Management consulting clarifies a decision: it analyses, builds scenarios and writes a recommendation. Business development consulting works as the owner of the problem and the result: it diagnoses, writes the plan, builds the implementation as well, and measures the outcome.

The difference lies not in quality but in scope and ownership; all three have their place. On the AI side, it maps onto this article's main question: the direction decision belongs to management consulting, a chatbot or content automation usually to an agency, and building a measured process and handing it to the team to the business development side. Which problem each of the three models solves, and which company should go to which, I covered in detail in what a business building studio is.

Where does INDOLES sit in this picture?

INDOLES is on the boutique side and describes itself as a business building studio: in AI consulting it starts with a task inventory, builds the pilot itself, measures the result side by side with the current method, and hands the system over to the in-house team. Strategy and software sit under one roof; the team that writes the recommendation also writes the prototype.

Two packages with open prices form the entry point to this work. The Digital Transformation Audit runs three weeks at €5,500: 3-5 processes are reviewed on site, pilot candidates are ranked by return, and the report separates out which of them actually need AI. The AI Pilot runs six weeks at €15,000: it locks onto a single use case, starts from no more than two candidates, includes a two-week field test with real users, and hands over the source code in full ownership; a second scenario is priced as a separate pilot. Both are list prices excluding VAT; model usage fees, cloud infrastructure and tool licences sit outside them. How the price band is built, and how the lines outside the price are estimated, we set out separately in what AI consultancy costs.

Two more things, stated plainly. INDOLES is a reseller of the İKAS e-commerce platform, and that is our only commercial tie on the software side; we have no partnership or commission arrangement with any model, cloud or AI software provider we recommend in AI consulting, and a finding of "AI isn't needed here" counts as a valid output. What falls outside our scope: internal change management and HR consulting, training models from scratch and academic research, and legal assessment. If your programme needs one of these from under the same roof, the criteria above are pointing you to a large consultancy — and that is the right decision.

Conclusion: the one page to fill in before you decide

The answer to whether you want a large consultancy, a boutique AI agency or an in-house team lies not in the providers' presentations but in how you describe your own work. Once the description is right, one of the options drops out on its own, and a decision left between the other two usually points to a mixed model.

The test you can run today takes one page. Write four lines: is the unit of work a single process or a multi-unit programme; who will do the implementation; who will run the system when the project ends; does regulation or the board require an independent signature? If you wrote "programme" on the first line and "yes" on the fourth, talk to a large consultancy. If you wrote "single process" on the first and "outside team" on the second, talk to a boutique team. If there is no name to write on the third line, whichever side you pick, your first job is to find that name.

Apply the same criteria to us. How the work was measured is in the Meccanotecnica Umbra case; the method's steps and what falls outside the scope are written out on the AI consulting page.