What Does an AI Consultant Do? A Real Week.

What an AI consultant does is close the gap between what AI can do for your business and what your team can actually make happen. That sounds abstract, so here is the concrete version: we spend our weeks talking to clients who are paying £4,100 a month across 23 separate software subscriptions for a 12-person company, and we help them figure out which three tools actually matter and how to connect them. We audit workflows, identify where time is leaking, build automations that work inside the tools the team already uses, and measure whether any of it sticks. According to McKinsey, only 11 percent of companies that attempt AI adoption see sustained productivity gains. The reason is almost always execution, not technology. An AI consultant's job is to be the execution layer, not just the ideas layer. If you want the fuller picture of what this role covers, what is an AI consultant is a good starting point.

A week in our calendar

The best way to answer what an AI consultant does is to walk through an actual week. Not a case study from 2019. A real one.

Monday: audit and baseline

Monday is usually diagnostic. We are inside a client's systems, not talking about them. For a 14-staff Dubai stem cell clinic, Monday meant logging into their CRM, their WhatsApp Business account, and their Bookimed portal to understand why 4 new patient bookings a month was the ceiling. We mapped every step between a patient's first WhatsApp message and a confirmed appointment. There were 11 manual handoffs. On average, response time was 38 hours.

By end of Monday we had a baseline: 4 bookings a month, 11 handoffs, 38 hours to first reply, £6,200 monthly spend on Bookimed referral fees.

Tuesday and Wednesday: build, not slide

No decks. We build on Tuesday and Wednesday. For the clinic, that meant configuring a WhatsApp qualifier flow: three questions (treatment, timeline, budget) that ran automatically before a human got involved. The qualifier connected to their existing booking system via a webhook. No new software. No 6-month integration project.

Wednesday afternoon is usually the first live test. We run it on real traffic, watch what breaks, fix it before end of day. For the clinic, the first version had a language issue (Arabic patients were getting English prompts). Fixed in 90 minutes.

Thursday: client handover and training

Thursday is the client's day. We do not build systems the client cannot run. If the operations manager cannot restart the flow after we leave, it is not done. For the clinic, Thursday was two hours with the front desk team: here is what the bot does, here is when it hands off to you, here is what to do if it gets stuck.

This is where most AI projects fail. The consultant disappears after launch. We do not.

Friday: measurement

By Friday of week 4, the clinic's numbers had moved: bookings up from 4 to 17 per month. Bookimed fees down 60 percent because more patients were coming through direct channels. £10,500 engagement cost. Net saving over the quarter was approximately £42,000.

Friday is also when we check in with other active clients. A London hospitality group running 8 venues and 22 staff had a different problem: their Gmail enquiry inbox was turning into a black hole. Reply time was 38 hours. Conversion rate on enquiries was 31 percent. We spent two Fridays calibrating their AI responder before it went live. After launch, reply time dropped to 12 minutes and conversion climbed to 58 percent.

What does an AI consultant actually charge?

Pricing is one of the least transparent things in this industry, so we will be direct. At twohundred.ai, our tiers are:

For comparison, a full-time AI lead in London costs £180,000 to £250,000 loaded (salary, NI, benefits, onboarding). A traditional agency retainer at the same budget often breaks down as roughly 40 percent overhead, 30 percent sales commission, 20 percent account manager, and about 10 percent on the actual work.

Fractional engagement means you get senior-level execution without the overhead. As an AI consultant for small business, the economics only work if the cost is proportional to what a 10 to 30 person company can absorb and the return is visible inside 60 days.

Is an AI consultant right for every business?

No. And we say that up front.

If you are below £500,000 in revenue, the workflow improvements are too small to justify the cost. If your team of 5 is already stretched and cannot commit 3 hours a week to testing new tools, the engagement will stall. And if the owner wants a dashboard that proves AI is happening rather than a process that makes money, we are probably not the right fit.

Who it does work for

The clients who see the sharpest results share three traits: they have a clear bottleneck (not a vague feeling that they "need AI"), they have someone internally who will own the new process, and they are willing to run experiments rather than demand certainty upfront.

8 of our 12 active clients saw a 4x improvement in qualified inquiry volume within 60 days. The common thread was not industry. It was having one operational problem specific enough to measure.

What happens after month one

Month one is the audit and the first build. Month two is refinement and second workflow. Month three is usually when the client starts spotting their own opportunities. That is the goal: by month three, we should be accelerating things they have identified, not hunting for problems ourselves.

A Manchester recruitment firm with 9 consultants is a good example. They had a Salesforce data problem: consultants were logging placements inconsistently, which meant their pipeline data was unreliable. We built a sync layer between their email and Salesforce. In 90 days, they recovered 22 placements that had been incorrectly marked as lost. That was approximately £160,000 in revenue that already existed, just invisible. Cost of the engagement: £10,500.

For a deeper look at how we structure this kind of work, see our AI strategy consultant page.

Frequently asked questions

What does an AI consultant actually deliver, week to week?

Concretely: audits of existing workflows, identification of automation candidates, configuration or build of automations inside tools you already use, training of your team on those tools, and measurement of outcomes. What they do not deliver is strategy decks that gather dust. If a consultant is spending most of their time in slides, that is a signal.

How is this different from a software agency or IT consultancy?

A software agency builds what you specify. An IT consultancy manages existing infrastructure. An AI consultant starts with a business problem and works backwards to the technology. The output might be a workflow automation, a data connection between two systems, or simply a decision not to use a particular AI tool. We are not incentivised to sell software. We are incentivised to make the metric move.

How long does a typical engagement take to show results?

For workflow automations (intake forms, responders, routing logic), results are usually visible in 30 to 45 days. For data or integration work (CRM cleanup, pipeline syncs), it is more often 60 to 90 days before the numbers reflect the change. We scope every engagement with an explicit "what does success look like at 60 days" before starting.

What if our team is not technical?

Good. Most of our clients are not. The clinic front desk team that learned to manage the WhatsApp qualifier had no technical background. We build for the person who will actually run it, not for the person who commissioned it. If it requires a developer to operate, it is not done.

Want this built for your business? Book a call.