AI agency vs AI consultant: which do you need?

An AI agency and an AI consultant are not interchangeable, and confusing the two is the most common reason businesses waste their first AI budget. An AI agency is a team that builds, runs, or manages AI systems on your behalf, usually under a project fee or retainer. An AI consultant is an individual, or a small firm acting as one, who diagnoses your situation, recommends a direction, and hands over a plan for someone else to execute. The AI agency vs AI consultant decision comes down to a single question: do you already know what to build, or do you still need someone to tell you? If you need a working system in the next 90 days, you almost certainly need an agency. If you need someone to decide which system is worth building before you commit budget, a consultant is the right first call.

What does an AI agency actually deliver?

An AI agency delivers working AI systems, not advice about them. The deliverable might be a customer service chatbot integrated with your helpdesk, an automated lead scoring model wired into your CRM, a document processing pipeline that replaces a manual review step, or a custom reporting layer that surfaces data your team currently extracts by hand. The agency owns the build end to end: scoping, architecture, integration, testing, and handover. You get a thing that runs, not a document about why a thing should run. Agencies typically charge a project fee for defined builds, ranging from around £8,000 for a narrow integration to £60,000 or more for a multi-system rollout, with retainers layered on top when ongoing operation is part of the deal. The distinction that matters most for buyers is whether the agency also operates what it builds or hands it over and disappears. Operator-model agencies stay embedded. Most do not. For the broader category definition, see what an AI agency does.

What does an AI consultant actually deliver?

An AI consultant delivers a diagnosis and a recommendation. The engagement usually opens with a discovery process: the consultant interviews your team, maps your current tools and workflows, identifies where AI would generate measurable return, and produces a written output. That output might be a strategy document, an implementation roadmap, a vendor shortlist, or some combination. Depth varies enormously and price is the clearest signal of it. A consultant charging £500/day for two weeks is likely producing a high-level report you could have outlined yourself. One charging £1,200/day for six weeks is probably doing a genuine audit: process mapping, data readiness assessment, and build cost estimation grounded in what your systems actually allow. Neither is wrong for every buyer. The only question is whether you need the diagnosis or the execution. If you want the full definition of the role, this builds on what an AI consultant is, and you can see how AI strategy consultants frame the same decision from the advisory side.

How do they differ in practice?

The two models differ in what they are accountable for. An agency is on the hook for working software and a delivery timeline. A consultant is on the hook for the quality of the thinking. That single difference drives almost everything else: team size, fee structure, how they sell, and what happens when something breaks.

What an AI agency does

• Builds and deploys AI systems such as chatbots, pipelines, automations, and integrations
• Owns the technical execution: architecture, coding, testing, deployment
• Works to a project brief or an ongoing retainer
• Stays accountable for delivery timelines and systems that actually run
• Ranges from 3 people to 300 or more, with overhead that scales with headcount

What an AI consultant does

• Diagnoses business problems and maps where AI genuinely applies
• Produces strategic recommendations, roadmaps, and vendor guidance
• Works to a time-based or deliverables-based fee
• Stays accountable for the quality of the analysis, not the build
• Usually a solo operator or a small team of 2 to 5 people

When does an AI agency make more sense?

An AI agency makes more sense when you already know what you want to build, you have a budget for execution, and you need a team to own delivery. If you have worked out that your customer service queue is the bottleneck and you want a triage chatbot connected to your helpdesk software, you do not need a consultant to confirm it. You need a team that can build it. The same applies if you tried to build something internally and stalled, if you have a specific process that needs AI support and no internal developer capable of doing it, or if you are running a pilot for a board that wants a working system this quarter rather than a strategy deck. Agencies are also the better call when speed is the constraint. The "$3,500/month for local SEO and I don't have 12 months to find out if it works" complaint that shows up constantly in business owner communities applies just as well to AI retainers. If you are paying every month, you want something running. For how agency fees are structured at each tier, AI agency pricing covers the full breakdown.

When does an AI consultant make more sense?

An AI consultant makes more sense when you do not yet know what to build, when internal stakeholders disagree about direction, or when the cost of building the wrong thing is higher than the cost of pausing to get the diagnosis right. A well-run consulting engagement saves money by narrowing scope before anyone writes code. If your business has 14 processes that could theoretically use AI and a budget that supports building three, the consultant's job is to rank those 14 by return on effort and steer you toward the three that pay back fastest. That is worth buying before you commit to an agency. Consultants also earn their fee when you are weighing build versus buy on a specific capability, or when the problem is genuinely ambiguous and rushing to execution would be expensive. The clearest consulting-first signal is an executive team that cannot agree on where AI belongs in their business. That conversation needs a neutral party with domain knowledge, not a build team that will optimize for whatever scope it is handed.

What about a hybrid: the embedded operator model?

Most SMEs do not fit cleanly into either bucket. They need someone who can diagnose the situation and then stay to execute the recommendation, rather than handing over a strategy document and walking away. This is the embedded operator model: a single point of accountability that combines the strategic framing of a consultant with the delivery responsibility of an agency. Instead of running a separate consulting phase and then a separate agency phase, each with its own brief, transition cost, and knowledge loss, an embedded partner runs them back to back with no gap. The embedded approach also removes the structural weakness of the pure agency model. Agencies build to the scope they are given, not to the outcome you actually need. If the brief is wrong, a pure agency builds the wrong thing precisely and on time. An embedded operator-partner pushes back when the data points to a different intervention. For how this compares to a technical hire, AI agency vs fractional CTO covers that decision in detail.

How twohundred approaches the choice

In practice, the agency-or-consultant question is usually a false binary, and treating it as one is what costs operators money. At twohundred, the approach is to assess first, then scope a working pilot, build it, and stay to iterate, so the diagnosis and the execution sit with the same team. That removes the handoff where a consultant's understanding gets compressed into a brief and an agency rebuilds the wrong model of your business from scratch. The advice for any buyer is simple: do not pay separately for thinking and building unless the thinking phase is genuinely independent of who executes. If it is not, find a partner who can do both and hold them to a working pilot, not a slide deck. You can see how that engagement model is structured on the AI implementation services page.

How do you evaluate either one?

The evaluation criteria differ, but the goal is the same: separate the partners who will own an outcome from the ones who will own a deliverable and then leave.

Questions for any AI agency

• What are the last three things you built, and can I speak to those clients?
• Who owns the systems you build after the engagement ends?
• Do you white-label tools, or build on what we already use?
• What happens if the build does not perform as expected within 60 days?
• What is the project fee versus the retainer component?

Questions for any AI consultant

• What does your deliverable look like at the end of the engagement?
• Have you personally implemented what you recommend, or do you refer it out?
• Can I see an anonymised example of output from a similar engagement?
• If you find that I need an agency, will you help select and brief one?

The clearest signal that you are talking to the right partner, agency or consultant, is whether they ask hard questions before proposing anything. An agency that quotes without a discovery conversation is optimizing for the sale. A consultant who produces a report without challenging your initial framing is optimizing for billable days. Both are the same failure wearing different clothes.

Frequently asked questions

Is an AI agency more expensive than an AI consultant?

Not necessarily, and comparing them on day rate misses the point entirely. A consultant who charges £1,000/day for four weeks (£20,000) and hands over a roadmap costs less up front than an agency charging £35,000 to build a working system. But the working system generates return and the roadmap does not on its own. Total cost of ownership, including the execution you will eventually pay for anyway, usually makes the agency model cheaper once you already know what to build. If you do not know, the consulting phase that stops you building the wrong thing pays for itself quickly.

Can an AI consultant also build?

Some can, and those are the most valuable engagements you can buy. A consultant who has personally built AI systems holds a fundamentally different view of what is feasible, what takes six weeks versus six months, and where the real integration complexity hides. Advisory-only consultants who have never built anything tend to produce roadmaps that look clean on paper and fall apart during implementation. If a consultant cannot point you to something they shipped into production, treat their timeline estimates with scepticism.

What engagement length should I expect?

Consulting engagements for AI strategy typically run two to six weeks for the initial audit phase, with a written output at the end. Agency builds for a defined scope usually run eight to fourteen weeks from brief to handover. Embedded operator engagements start with a two-to-four-week diagnostic and then roll into a build phase. Be suspicious of any engagement that asks you to commit beyond six months before you have a working pilot in front of you.

Do I need both a consultant and an agency?

Sometimes, but far less often than vendors would like you to believe. If you hire a consultant and then separately hire an agency to execute, you pay twice for the knowledge transfer that happens between them. The brief the agency receives is a filtered summary of what the consultant learned, and information degrades in that handoff. If the consultant and the agency are the same party, or the consultant stays involved through execution, that degradation never happens, and you stop paying for the same understanding twice.

What is the fastest path to a working AI system?

A narrow scope with a clear success metric, a partner who builds on the tools you already use, and a decision-maker who can approve the pilot without committee sign-off. The projects that drag on for eighteen months are rarely complex. They are under-scoped in week one and over-governed in weeks two through sixteen. Pick one painful, well-understood process, define what success looks like in a number, and build that before you build anything else.

---

Related Services

For businesses that want strategic guidance on where AI fits before engaging a vendor, AI consulting services explains the advisory approach. Hands-on deployment is covered in AI implementation services.