I help you pick the AI bets worth making, then tell you straight whether each one paid off.
I've spent 30 years founding and running businesses, seven of them my own, and these days I run my own on AI. I start with you and your leadership team using it in your own work. Then we get one real improvement working with your people and measure what it's worth.
Email me to arrange a coffee or a 25-minute call · james@jamesvarga.com
“Customers buy from people, so I'd keep your people in front of them and use AI on everything behind. If the plan is cutting headcount, I'm not the right fit.”
The lesson I'd pass on is that agents are only as good as what they know. Giving them a shared picture of the business, and a clear point where a person decides, made the difference.
Over coffee I'll show you the ten-minute version, live, including the morning briefing I start every day with.
If two or three of these sound familiar, someone may already own AI on paper. What's usually missing is the time and hands-on experience to turn it into changes people actually use. That's the job I take on with you.
So who's joining up your commercial priorities with the changes your people need to make?
A lot of the AI pitch you'll hear is about taking people out of sales and customer service. I think that's the wrong end of the business. Your customers buy from people, and those conversations are where trust and repeat business come from.
The bigger opportunity is inside. It's the reporting, chasing, drafting and searching that eat your people's week. Hand that to AI and the account manager who loses Friday to reports spends it with clients instead.
The aim is 100 people who perform like 140.
In our first conversation I'll ask you to mark on one page where your business is, then where each of your leaders is. Try it here. Point at a step and the puck moves to it.
Most businesses I meet are running ahead of their leaders. People take their lead from their boss, so AI only spreads where each leader uses it in their own work. That's why I start with you and your leadership team.
None of this is magic, and you'll see all of it. Six things you can hold me to.
Each AI bet gets a card in a decision record. It names the problem, where we're starting from, who owns it, what it should return and what it costs, and why we went ahead or stopped.
Within a day of our first conversation you'll have a one-page note on where I'd start and what we'd need to check first. In month one we confirm the owner, the starting figures, the test it has to pass and the date we decide.
Week one includes a leadership AI day, and each leader leaves with AI working on one of their own tasks. After that I turn up each month with context and follow through between meetings.
Five measures, reported separately, so a good number in one never hides a bad one in another. More on that below.
A place at the quarterly CEO roundtable.
Where it helps the chosen work, I put what your business knows in one place AI can use. It lives in your own accounts and stays yours.
One decision to start. We agree one sponsor, one owner, one problem and one team to begin with, and the wider leadership work starts straight away.
Each month is a step up from the last, and at the end of month three you decide whether to carry on. Point at a step to see what happens in it.
A one-page plan, and every leader with AI working on one of their own tasks.
It's in real use, and the evidence is building.
Your call on whether to continue, change or stop, and the next quarter's priorities.
Once the first improvement is working, the job changes. I keep what we've built useful as your tools and your business change. I help the owners clear whatever is holding them up, and we pick the next improvement worth the money. When a supplier pitches you AI, you'll have someone to tell you whether it's worth having.
We meet for two working sessions a month, with the building and follow-up done between them. You get a short monthly account of results and the decisions needed, and the decision record stays current. Each quarter, more of the work moves to your own people. It's planned improvement work, not an on-call support desk.
Monthly, with no minimum and 30 days' notice.
Leadership AI day. Every leader leaves using AI on one of their own tasks. This is when you get most of my time.
First improvement live. I build it with the people who'll use it, so they learn by doing rather than watching.
Your call. Continue, change or stop, on the evidence against where we started. Nothing rolls on by default.
Your team leads. I step back to two or three days a month, then lighter, as your owners run it and bring me the next problem.
Why now Awaab's Law puts hard deadlines on damp and mould, and the regulator wants the evidence.
Reports, inspections and letters gathered into one record, with deadlines tracked and the evidence collected as the work happens.
We'd measure cases closed inside the deadline and the time spent pulling evidence together for the assurance statement.
We'd start with one team's cases and the housing system as it is, with no data clean-up first.
Why now Anti-money-laundering supervision is moving to the FCA, and consolidation and partner succession put what the firm knows at risk.
First drafts of proposals and engagement letters built from the firm's own past work, ready for a partner to review.
We'd measure hours per proposal and admin time per fee earner.
We'd start with fixed-fee and non-billable work, so the time saved doesn't cut fees.
Why now Experienced people are leaving faster than they can be replaced, and quality systems need evidence.
Procedures, fault histories and job notes turned into answers the whole team can ask, each one cited to its source.
We'd measure how long new starters take to work unsupervised, and repeat faults.
We'd start with one line or product family, alongside the engineers who know it best.
Why now Tighter rules for higher-risk buildings demand documented compliance, on margins that leave no room for unrecovered work.
Instructions and changes checked against the contract as they arrive, and compliance evidence filed as the work is done.
We'd measure unrecovered variations and the hours spent on handover documents.
We'd start on live projects rather than the archive.
Why now Consumer Duty and operational resilience rules mean every decision has to be explainable.
Policies, procedures and case notes your staff can ask, with cited answers, so a reviewer can see why a decision was made.
We'd measure file-review time and rework on complaints.
It runs where your risk team is comfortable, with access agreed before anything goes in.
Why now Customers expect a price the same day, and margins are thin.
Quotes drafted from past jobs and current prices, for the branch team to check and send.
We'd measure quote turnaround and the share of quotes won.
If the ordering system's data needs tidying up, we'd pick a first improvement that doesn't depend on it.
These are illustrations of how I'd approach each sector, not client case studies.
I report results as five separate measures. Where the evidence supports it, we also work out the overall return after costs, without counting the same benefit twice. If something isn't earning its place, I'll recommend we stop it.
I measure the leadership side too. That means how many of your leaders use AI in their own work each week, and how many have a change running in their area, with a short confidence check each quarter.
| I own | You commit to |
|---|---|
| Agreeing priorities with you | A senior sponsor |
| Making sure each piece of work has an owner inside the business | Someone to own the first improvement |
| Coordinating the delivery, and any specialists | Access to the information and systems it needs |
| Raising problems and decisions promptly | Making decisions in good time |
| Measuring whether each change is working | Time for your people to try it and say what they think |
Headcount and sector matter less than these. It works best when most of them are true.
You've got choices, and some of them are good ones. Here's how I'd think about each.
| Option | What you get | Where I fit |
|---|---|---|
| Your IT provider | Systems access and trust, with AI as an add-on | Keep them on your infrastructure. I own the commercial priorities and getting people to use it. |
| A keen manager and a course | Enthusiasm and know-how, without senior ownership | I give that manager a senior partner and a plan your leadership team owns. |
| A Head of AI | One dedicated owner, at £100,000 or more, after a 3 to 9 month search | Hire one when your leaders can direct one, and I'll help you hire well. |
| A fractional Chief AI Officer | Senior advice, and some build too | I've run the business, and I start with your leaders using it. |
| A packaged service or a one-off build | A defined tool or automation, cheaply | I'll help you pick which one is worth building. |
| A CEO peer group or a business school course | Peers or knowledge | Done on your own business with your team, and I stay on. |
| Waiting | Nothing spent yet | What is it costing you while it stays patchy? |
No cost, and no follow-up unless you ask for one.
Eight to twelve CEOs around one question, what's actually paying off? Chatham House rule, and no pitches.
Ask for an invitationForward me an AI pitch, a proposal or a question from your board. You'll have my view in three lines within a working day.
Send it to meWhat's real in AI for a business like yours this month, and what you can safely ignore.
Ask to be addedSomeone who has run a business, working alongside you and your leadership team to pick the AI bets worth making and get them working with your people. It starts with a 90-day opening and carries on monthly for as long as it's useful.
Some fractional Chief AI Officers focus on strategy and governance, and some build too. I've run businesses myself, so I start with you and your leadership team using AI in your own work. Then I build the first improvement alongside your people and keep it running.
A senior AI lead can help, but one person can't carry it if the leadership team isn't using AI themselves. Hire one when your leaders can direct one. When the time's right, I'll help you hire well.
That isn't what I'm for. The aim is the same team doing more of the work only people can do, with AI taking the rest. If the plan is mainly headcount reduction, I'll say so in our first conversation and point you elsewhere.
You can stop at the end of any month. The partnership after the first 90 days is monthly too, with no minimum and 30 days' notice.
You do. Everything is built in your own accounts, so your data and the work stay yours if we stop working together. I'll sign an NDA before we start.
No. You need one person who can own the first improvement, and access to the systems it touches. If you have an IT provider, they keep looking after your infrastructure and sign off new tools.
Yes, if you'd prefer. There's a version with a lower monthly fee and a capped payment tied to one measured improvement. We agree the starting figures, the evidence and the cap before we begin.
Only if we agree it first. Software, travel, specialist help and bigger builds are quoted separately, before anything is spent.
I do. I lead every engagement and do the leadership work myself. If a piece needs a specialist, I'll bring one in once we've agreed it's needed, and they contract with you directly.
Send me a line about your business and where AI is or isn't working. Over coffee, or a 25-minute call if we haven't met, I'll mostly ask about you. I'll show you my own set-up on a problem you've just named, and if it isn't a fit, I'll say so.
Within a day you'll have a one-page note on where I'd start and what we'd need to check first, in a clean copy you can pass to your chair or FD.
If you're reading this with someone else in mind, it suits an established business of typically 50 to 250 people where AI is patchy and the CEO can give it some time. Send them the link. I'll offer them one conversation and won't chase.