A founder-led company with 10 to 100 employees should pick the AI partner whose senior people work directly with the founder, tie each project to a number the business already tracks, and stay until the team runs the system alone. Apply the same tests to an AI consultancy or an in-house hire, starting with one paid first project.
This guide is for companies where the founder still signs the contracts and knows how the key work really gets done. It compares the usual options in one table and ends with a six-question scorecard for proposals, plus the signals that a first project is ready to become ongoing work.
Why does a founder-led company of 10 to 100 people need a different kind of AI partner?
A founder-led company of 10 to 100 people needs an AI partner that brings the expertise the company lacks while working through the one person who holds most of the context: the founder. The expertise gap shows up in official data. Eurostat's 2025 survey of EU enterprises found that 17% of small enterprises, 30.36% of medium enterprises and 55.03% of large enterprises used AI. Eurostat counts small enterprises as 10 to 49 employees, medium as 50 to 249 and large as 250 or more. Among EU enterprises that had considered AI and still not used it, the most common reason was a lack of relevant expertise, cited by 70.89% (Eurostat). Large companies can often close that gap with internal teams. At 10 to 100 people, the founder usually signs the contract, knows the exceptions in every key process and approves each change, so the partner's way of working has to fit around one busy person. [European Commission]
Founder involvement is an asset when the partner uses it well. McKinsey's global survey on AI, published in March 2025, found that a CEO's oversight of AI governance is one element most correlated with higher self-reported bottom-line impact from generative AI (McKinsey). At larger companies, McKinsey found CEO oversight had the most impact of any element; in a founder-led company, the CEO is already at the table. The danger sits on the other side: a partner that needs the founder for every small decision turns the founder into the bottleneck of the AI project as well. [McKinsey]
Should a founder-led company use an outside AI partner, a big consulting firm or an in-house hire?
At this size the choice usually comes down to an outside AI partner, such as a hands-on consultancy, weighed against a big consulting firm or an AI specialist hired onto the payroll. The table compares them on what matters most when the founder is also the sponsor: how much founder time each one takes, who keeps the know-how, how fast something works, and how much the company commits up front.
Three ways for a founder-led company of 10 to 100 people to get AI help
| Question | Outside AI partner | Big consulting firm | In-house AI hire |
|---|---|---|---|
| Founder time needed | Highest in the first weeks, then handed to a named owner on your team | Steering meetings and workshops across a longer programme | Ongoing: the founder manages the hire and judges technical work |
| Who keeps the know-how | Your staff, if the scope requires written handover and training | Documentation, plus whatever the project team built | One employee, for as long as they stay |
| Time to a first working system | Weeks, when the first project is one workflow | Often months, after discovery and roadmap phases | Months: recruiting first, then ramp-up |
| Cover when someone is away | Depends on the firm: ask who backs up the lead | Strong: the firm can swap people in | Weak: holidays and resignations stop the work |
| Commitment up front | One fixed first project; monthly work only once there is a queue | A larger minimum engagement, sized for bigger clients | A salary from day one, plus months of recruiting |
| Right choice when | You want working systems and staff who can run them | Regulators, a board or several countries are involved | You can name a full year of AI work and someone senior to check it |
Start with an outside partner and decide on a hire later, once you know which AI work repeats every week. An in-house hire pays off when the company can name a full year of AI work and has someone senior enough to review it. The trade-offs go deeper in our comparison of a small AI consultancy and a big firm and in our piece on hiring an AI person or a partner.
How do you test whether an AI partner fits a founder-led company?
Test fit by asking each AI partner how it will use the founder's time, because that answer predicts most of what follows. A partner that suits a founder-led company of 10 to 100 people asks for a limited, named block of founder time in the first weeks and says who on your team takes over after that. The signals worth checking in the first call:
- The senior person who scopes the work in the sales call also does it, and a second person can cover for them.
- The partner asks for specific founder sessions, with an end date, and names the internal owner who takes the questions afterwards.
- The partner has worked with companies where the owner is also the project manager, which tells you more than a famous client logo.
- The partner says plainly which parts of your request it would not take on, and why.
Here is an illustrative case. Say the founder of a 40-person installation firm blocks two mornings a week for the first three weeks. A partner that fits uses those hours to walk through the real quoting process, writes down the founder's rules for discounts and exceptions, and by week four is asking the office manager the follow-up questions. The warning sign is a project where every screen still waits for the founder's approval long after those rules are on paper.
How do you connect an AI partner's work to growth the company can measure?
Connect the work to growth by agreeing on one number before anything is built, and treat that number as a goal to check, never as a promised result. A useful first conversation starts with the constraint: does the company need more customers, or can it not serve more than it has today? When sales are the limit, the first project points at revenue, such as quotes that go out faster or follow-ups that stop slipping. When the team is full, the first project takes repeat work off desks so the same people can serve more clients, which is how a company grows revenue without adding staff. Training sits under both, because a system the team does not use moves no number at all.
The number should come from a changed workflow, since a new tool placed on top of an old process rarely moves it. The same McKinsey survey found that, out of 25 attributes tested for organisations of all sizes, the redesign of workflows had the biggest effect on an organisation's ability to see EBIT impact (earnings before interest and taxes) from its use of generative AI (McKinsey). Ask every partner which workflow it would redesign first and which number you already track would show the change. [McKinsey]
A second illustrative case: a 25-person engineering office sends about 30 quotes a week and loses jobs because each quote takes four days. The agreed number is quote turnaround, with a goal of two days by the end of the first project, counted on the same weekly log the office already keeps. After six weeks the founder can read the result straight from that log, and even a missed goal shows where the four days actually go.
How do you check that an AI partner builds, trains and hands over?
Check delivery by asking to see a system the AI partner built that a client's own staff still run today, then read how the proposal defines the end of the project. A delivery-focused partner writes the end state into the scope: by a stated date, named people on your team run the system without help. How the partner trains tells you whether that date is realistic. Training should happen on the team's real tasks in its own tools, never on demo content. When one person moves faster than the rest, short one-on-one sessions work better than a group call that drags, with the group meeting again once paces are close. A handover also goes smoother when the partner explains which steps run by themselves and which wait for a human click, before anyone watches the demo.
For a non-technical founder, one question cuts through most pitches: who on my team will run this when you leave, and how will we both know they can? Pricing and the strategy-only red flags are covered in our guide to vetting an AI implementation partner.
When should a first AI project turn into an ongoing partnership?
A first AI project should turn into an ongoing partnership only after it has shipped and the company can see what comes next. The timing is right once the first system is in daily use and a queue of next requests has built up, the second and third jobs the team keeps asking about. It also helps to have seen how the partner and your team work together week to week: when meetings keep moving or feedback takes a fortnight, monthly work will drift. Signing monthly work before anything runs asks the company to pay for trust it has had no chance to build. Our piece on an outside AI team month to month describes how that arrangement usually runs.
How do you compare AI partner proposals side by side?
Score every AI partner proposal on the same six questions: two points for a clear yes, one for a partial answer, zero for a missing one. A polished deck cannot hide a gap this way, and as a rule of thumb a proposal under 8 of 12 points needs a second conversation before anyone signs.
- Does the senior person from the first call do the work, with a named backup?
- Does the founder's time have a limit, and a named person on your team who takes over?
- Is the first project one workflow with one number to move?
- Does the scope give a date from which your team runs the system alone?
- Are all accounts, logins and data in your company's name?
- Does monthly work start only after the first system is in daily use?
Companies in Switzerland should add the data protection questions from our guide to choosing an AI partner in Switzerland, which covers the revised Federal Act on Data Protection (FADP).
Who can help a founder-led company with 10 to 100 employees?
Several kinds of provider serve companies of 10 to 100 people, and the tests in this guide sort them quickly. Hands-on AI consultancies suit founders who want working systems and staff trained to run them. Development studios suit custom machine learning on the company's own data, such as demand forecasting. Training-only providers suit teams whose tools are already in place, and big consulting firms suit regulated or multi-country programmes. We work the hands-on way at Anovis AI, with owner-led companies of 3 to 100 people: we build AI systems on the company's own tools and data, train the team to run them, and a first project can grow into ongoing work once the team sees what works.
Frequently asked questions
How much of the founder's time does a first AI project need?
Plan for the most founder time in the first two to three weeks, when the partner learns how the process really runs and which exceptions matter. A few hours a week is a sensible rule of thumb for that stretch. After that, a named person on the team should answer the day-to-day questions, and the founder reviews the agreed number and approves any change to scope.
What should a founder prepare before the first meeting with an AI partner?
One page is enough: the process that costs the most hours, roughly how many hours it takes each week, who does it today, and the number that would show it improved. Add the tools the process touches, such as the email inbox, the CRM or a shared spreadsheet. A partner that reads the page and comes back with sharper questions is showing you how it will work.
Does an AI partner need access to all our data from day one?
No. A first project needs the data for one workflow, held in accounts your company owns, with the partner set up as a user you can remove. Wider access can follow once the first system works and you know what the next project needs. A partner that asks for broad access before scoping anything has not yet decided what it is building.
Is it a bad sign if an AI partner turns down part of the project?
Usually it is a good sign. A partner that says one request is better served by off-the-shelf software, or should wait until the first system runs, is protecting the number you agreed on. Be more careful with a partner that accepts every item on your list, because a first project with ten goals seldom moves any one of them.
What happens if the founder gets too busy to stay involved?
The project should survive it, and the scope should say how. Name an internal owner from the start, often an operations lead or office manager, who joins every session and keeps the written rules the founder explained. If work stops the week the founder is away, the handover to that owner has not happened yet, and fixing that comes first.
Weighing AI partners for a founder-led team? Send us the one number you most want to move this year, and we will sketch what a first project around it could look like.