TL;DR: Most advice firms now have an AI pilot, but few see it change the business. The firms that do have sorted out three things around the model: connected, clean data, a team trained in AI like any other professional skill, and a culture where leaders use it visibly and measure the small wins.

At nearly every event I've spoken at this autumn, someone puts their hand up and asks some version of: "Why is your model better than your competitors'?"

It's a fair question, but I've come to think it's the wrong first question. The firms getting real value from AI didn't pick the latest, cleverest or most expensive model. They sorted out three far less glamorous things: their data and integrations, their team's AI skills, and their culture.

The Customer Engine Problem

In Formula 1, several teams run the same engine. Same horsepower, same supplier, same spec. And yet every race they finish miles apart on the grid. What separates them is everything around the engine: the chassis, the data, the pit crew, the driver, and the way the team works together.

AI in advice is heading the same way. The underlying models are improving quickly, and over time most providers will have access to very capable ones. What won't be handed to you is the stuff around the engine. That's where the gap between firms is opening up, and it's opening up fast.

Softwire's research found 98% of UK asset and wealth management firms have an AI pilot live, but only 12% say it's transforming their business. Plenty of firms have bought the engine. Far fewer have built the car around it.

There's a second reason the benefits stall. Using AI for one step of the client journey doesn't add capacity to the whole firm. AI note-takers speed up the front of the journey, but the queue simply re-forms further down the line with your paraplanners. Speed up one step on its own and the bottleneck just moves.

In one 10-adviser firm we work with, all 13 people on the platform use AI for meeting notes, but only two use it to draft suitability letters. The front of the journey got faster and the report-writing stage barely moved.

Compare that with Timothy James & Partners. They connected AI meeting notes into their annual review process, so their support team gets structured notes within minutes of a meeting ending instead of waiting for the adviser to finish them. Advisers, paraplanners and support now work in parallel rather than in sequence, and the firm has at least halved the time it takes to produce post-meeting notes and follow-ups. That takes connected data and people who are willing to redesign the whole process, so every link in the chain gains capacity together.

1. Data: You Can't Get Clever on a Data Swamp

The slide that gets the biggest reaction in our presentations is about data, and barely mentions AI. Putting AI on top of incorrect data just gets you to the wrong answer faster. We ask firms to consider where they are in terms of data maturity:

• Data silos: six to eight disconnected systems (back office system, cashflow tool, risk profiler, platform, fund research, client portal, vulnerability log and so on)

• Data swamp: everything pulled together, but unstructured, duplicated and inconsistent

• Data lake: cleaned and consistent

• Data warehouse: organised so it can be reused across the whole firm

AI only gets clever once it has at least a lake, ideally a warehouse. Most firms are still typing the same data into several systems, living with one-way integrations and adding yet another login to an already crowded stack.

Take a very simple example: me.

If my adviser records and transcribes our meeting, I'm Steve. If one of the team types my name into the back office system, I might be Steven. If I fill in a digital fact-find myself, I'm a Step-hen.

Multiply that by 500+ fields of data for each client, and hallucination stops being the main risk. The bigger problem is that the AI doesn't know which of the duplicate values to trust.

Your data doesn't need to be perfect before you start, and using AI well can help tidy it as you go. One Chartered Financial Planner told us that because this year's annual review notes all follow the same format, next year's preparation will be easier. But someone needs to own the journey from silos to warehouse, and it needs to be on the board agenda.

2. AI Capability: Firms Assess Everything Except This

Advice firms are brilliant at measuring competence. CPD logs, file reviews, qualifications: the industry has built an entire machine around making sure people know what they're doing.

And yet almost nobody assesses how capable their team is with AI, or invests in helping them get better. That still surprises me.

The Conference Board's 2026 research found more than half of workers now use AI daily or weekly, but only a third had taken part in any employer-provided AI training in the previous six months. Most people are teaching themselves, on the job, with whatever they've picked up online or heard from colleagues.

In a regulated business, that is a data protection problem as well as a productivity one. A 2026 Resume Now survey of US workers found 76% had used AI tools they found themselves for work, and separate research for PagerDuty found about a third of office professionals had entered customer data into public AI tools. If you don't give your team the right tools and the training to use them, many will find their own, and your client data goes with them.

That creates a pattern we see in almost every firm: an 80/20 split. A handful of enthusiasts racing ahead, and a majority who tried it once, got a slightly odd answer, and quietly went back to the old way. Meanwhile the way people work with AI has moved through three stages in three years, from simple prompting, to giving it proper context, to setting it goals to work towards, and most businesses are still on the first.

So, three practical suggestions:

• Baseline it. Run a simple capability assessment across the team. The point is to know where you're starting from, so nobody should feel tested.

• Invest in it. Budget for proper courses and protected time to practise, the same way you would for any other professional skill or qualification. Capability builds with use: among people who stay on our platform, the typical user produces three documents in their first month and eight by month six.

• Make it role-specific. What a paraplanner needs to get good at is different from an adviser, which is different again from your compliance team.

3. Culture: The Bit You Can't Copy

You can buy the same software as your competitor. You can hire the same consultants. You can, eventually, get your data into similar shape. What you can't do is copy their culture.

Culture is why two firms can roll out the same tool on the same day and get completely different results six months later. In one, the managing partner uses it in front of the team, shares what worked and what didn't, and makes it safe to experiment. In the other, it's announced by email, a few people log in, and it slowly becomes "that thing we pay for".

We've seen one firm live through both versions. A small advice firm first tried AI around two years ago and dropped it, because the team was growing too fast to implement it properly. When they came back, they wrote an AI policy first, told the team where the firm stood before choosing a tool, and gave the rollout three months. One of their paraplanners went from one suitability report a day to three.

What works, from what we've seen:

• Leaders go first, visibly. If the people at the top aren't using it, nobody else will believe it matters.

• Appoint champions. One person per team whose job includes helping colleagues, collecting wins and flagging problems. One consolidator we work with gave a senior paraplanner 18 hours a week to lead their rollout, with authority to make decisions for the team. Their withdrawal reports went from 45 to 60 minutes each to 10 to 15 minutes of editing, and a three to four week backlog cleared in a fortnight.

• Measure success and celebrate the small wins. Capture how long a task takes, and how many times a month one person, or the whole team, completes it. Very few firms do this today. When we spoke to seven firm leaders this summer about the time AI was saving them, every figure was an estimate made on the call, and none had a baseline to compare against. An hour saved on a suitability letter, told well in a team meeting, does more than any rollout plan.

• Give permission within guardrails. Be clear on what's allowed, then actively encourage people to try things.

A Five-Minute Health Check

If you want a quick read on where your firm stands, try answering these honestly:

1. Could you name every system your client data lives in, and does anyone own joining them up?

2. Do you know how capable each member of your team is with AI today?

3. Have you invested in any AI training in the last six months?

4. Who are your AI champions, and do they know they are?

5. Does your leadership team use AI themselves, in a way the rest of the firm can see?

If you're scratching your head on more than a couple, there's no need to panic. Start there, before you start comparing models.

What Separates the Firms That Get There

The models will keep getting better whether you do anything or not. Your data, your people's capability and your culture won't. Those are the parts that take time, and they're what will separate the firms that tried AI from the firms it changed.

It won't only show up in your efficiency numbers either. Consolidators and private equity buyers are starting to look at how deeply AI is embedded in a firm, because it tells them how durable the productivity gains are and how easily the business will integrate after a deal. Your data, your people's capability and your culture are becoming part of what your firm is worth.

By all means ask your provider why their model is better. Just make sure you're asking yourself the harder questions too.