Where to Start with AI: A Guide for Mid-Market Leaders

The most common question mid-market business leaders ask about AI is not "should we use AI?" - they already know the answer is yes. The question is "where do we actually start?"

It's a harder question than it sounds. According to McKinsey's 2025 UK research, 73% of companies in the £20M–£150M range have experimented with AI, but only 22% have connected those experiments to a documented business strategy (McKinsey, 2025). Most companies have started - they've just started in the wrong place.

This guide is for leaders who want a practical, step-by-step answer. Not theory. Not hype. Just the sequence that works.

The Wrong Place to Start

Most companies start with AI in one of three ways. All three are wrong:

Starting with a tool. "Let's get ChatGPT Enterprise" or "We should look at Copilot." This is procurement, not strategy. Without knowing what business problem you're solving, any tool purchase is a gamble. According to MIT's 2025 research, purchasing vendor solutions has a 67% success rate - but only when connected to a clear business need. Buying a tool first and looking for a use case second inverts the logic.

Starting with a pilot. Running a small AI experiment sounds low-risk, but without strategic direction it creates pilot purgatory - where pilots work in isolation but never reach production scale. According to IDC's 2024 research, 88% of AI proofs of concept never make it to production.

Starting with training. "Let's upskill the team on AI" sounds proactive, but training without context is wasted effort. People complete the course and go back to working the way they always have, because nobody told them what to do differently.

The Right Place to Start: The Three-Step Foundation

Step 1: Understand where you are (Week 1-2)

Before making any AI investment, answer three questions honestly:

What AI is already in use? This is the shadow AI question. According to Gartner's 2025 research, 69% of organisations suspect or have evidence that employees use unauthorised AI tools at work. Your people are almost certainly using ChatGPT, Gemini, or similar tools - with company data, on personal accounts, without oversight. Understanding your shadow AI landscape is the essential first step because it reveals both the risks and the opportunities.

What does your data look like? AI is only as good as the data it works with. According to Gartner, 85% of AI projects fail due to poor data quality. Before buying any AI tool, assess whether your data is ready - is it clean, accessible, governed, and relevant to the problems you want to solve?

Is your leadership aligned? This is the most important question. According to EY's 2025 UK AI Barometer, only 26% of mid-sized UK businesses say their leadership is "fully aligned" on AI. If your CEO, CTO, and CFO each have a different vision of what AI should do for the business, no amount of technology will help. The AI advice gap must be closed before meaningful investment begins.

Step 2: Build the strategy (Weeks 3-6)

With a clear picture of where you are, build a strategy using the AI Strategy Compass - six components that every real AI strategy must address:

1. Business alignment - Which specific business outcomes will AI improve?
2. Prioritised roadmap - What do we do first, second, third?
3. Governance framework - How do we use AI safely and compliantly?
4. Capability plan - What skills do our people need?
5. Vendor assessment - Which tools fit our needs?
6. Measurement framework - How do we know if it's working?

The strategy doesn't need to be a 50-page document. A one-page plan that the leadership team can align around is more valuable than a detailed report that sits in a drawer. According to BCG and MIT Sloan Management Review's 2024 research, companies that align on strategy before investing see 3x higher returns and move 40% faster from pilot to production (BCG x MIT, 2024).

Step 3: Execute one initiative well (Weeks 7-12)

Don't try to do everything at once. Pick one initiative that:

• Connects to a measurable business outcome
• Uses data you already have (or can get quickly)
• Has a named business sponsor (not just IT)
• Can show results within 90 days
• Is visible enough to build internal confidence

The best first initiatives are often unglamorous: invoice processing, report generation, customer enquiry triage, data cleaning. According to MIT's 2025 research, over 50% of generative AI budgets go to sales and marketing, but the biggest ROI is in back-office automation.

Get one thing working. Measure the results. Then expand from there.

What This Looks Like in Practice

Week	Activity	Outcome
1-2	Shadow AI audit, data readiness assessment, leadership alignment conversations	Clear picture of current state
3-4	Strategy workshop with leadership team, prioritise use cases	One-page AI strategy aligned across leadership
5-6	Governance framework established, approved tools selected	Policies in place, shadow AI addressed
7-8	First initiative launched with clear success metrics	Pilot running with business sponsor
9-10	Monitor, measure, iterate	Early results visible
11-12	Go/no-go decision on scaling, plan second initiative	Evidence-based expansion or pivot

This 12-week framework gets a mid-market company from "where do we start?" to "here's what's working and here's what's next" - without wasting money on tools that don't fit or pilots that don't scale.

Who Should Lead This?

Only 9% of mid-market companies have a CAIO or equivalent role (Gartner, 2025). Most companies starting with AI don't have dedicated AI leadership - and they don't need to hire a full-time executive to get started.

Options:

Approach	Best for	Cost	Timeline
Internal champion	Companies with a strong CTO or COO who can dedicate time	Free (opportunity cost)	Slowest - learning while doing
AI consultancy	Companies that want a one-off report	£50K-£200K	Fast report, slow execution
Fractional CAIO	Companies that want embedded leadership and execution	£2,500–£7,500/month	Fastest path to results

A fractional CAIO provides the strategic direction, governance expertise, and leadership bandwidth that most mid-market companies lack - without the £250K-£400K cost of a full-time executive hire.

Common Mistakes When Starting with AI

Trying to boil the ocean. Companies that start 5 AI initiatives simultaneously usually finish none. Start with one. Get it right. Expand.

Delegating to IT. AI is a business transformation, not an IT project. According to Pertama Partners' 2026 research, 61% of failed AI projects were treated as IT initiatives rather than business transformation. The leadership team must own this.

Skipping governance. AI governance feels like bureaucracy when you're eager to start. It's not - it's risk management. Only 7% of UK businesses have proper governance. The other 93% are exposed.

Waiting for the "right" time. There is no perfect time to start with AI. The technology will keep changing, the market will keep moving, and your competitors won't wait. The best time to start is now, with a structured approach.

Letting vendors set the agenda. AI vendors will tell you their product is the solution to whatever problem you describe. Start with your own diagnosis before talking to vendors.

Frequently Asked Questions

How much should we budget for getting started with AI?

For the 12-week foundation described above, budget £10K-£30K including a fractional CAIO engagement or equivalent advisory support. This covers the diagnostic, strategy, governance setup, and first initiative. It's a fraction of what most companies waste on uncoordinated tool purchases - according to McKinsey, mid-market companies waste £200K–£2M across scattered AI initiatives.

Should we hire an AI team?

Not yet. Most mid-market companies don't need dedicated AI headcount to get started. They need strategic direction and governance - which a fractional CAIO or advisor can provide - plus upskilling of existing staff. Hiring comes later, once you know what roles you actually need.

What if we've already started and it's not working?

That's normal - more than 80% of AI projects don't deliver value (RAND, 2025). The fix is to stop, audit what you have, build the strategic foundation described above, and then decide which existing initiatives are worth continuing and which should be killed. A structured restart is better than continuing to invest in something that isn't working.

How long before we see ROI?

For well-chosen first initiatives with clear metrics: 3-6 months to measurable impact. The 12-week framework above is designed to get your first results within a quarter. Full AI maturity is a longer journey, but you should see evidence of value quickly.

What's the single most important thing to get right?

Leadership alignment. According to BCG and MIT's 2024 research, only 10% of companies generate significant financial value from AI, and the common trait among them is that leadership aligned on strategy before investing. Everything else - tools, data, training, governance - follows from that alignment.