Most firms I work with underprice their first engagement. They can do the maths. The problem is the pricing lives in a spreadsheet and somebody's memory, and neither of those is in the room when the client says yes.
I'm a CPA. I trained at KPMG, then built Quanto, where we now help accounting firms grow using the same AI tools we built for our own practice first. Pricing is where I see the most money left on the table, and it's the thing firms find hardest to fix on their own.
I built a pricing tool live in Claude during my session at the AI in Practice Summit. If you would rather watch the build than read about it, the recording is here. The build itself took a few minutes. Getting it to price like a partner I trust took everything that came before it.
Pricing is a lemonade stand. We just forget our own time.
Think about a lemonade stand. You have raw ingredients, lemon, water, sugar, ice. Then you have the input we always forget, your time. Spend two dollars on materials and a dollar of your time, sell at $2.50, and you have lost money on every cup.
A firm is the same, with higher stakes. Your inputs are your qualification, the years you spent earning it, your software, your team's wages and your time. The output is clean financials and a client who sleeps at night. We price our clients' businesses with that clarity every day. We rarely turn it on ourselves.
The result is an engagement priced at $500 a month that should have been $20,000 a year, and a relationship that becomes too busy to ever go back and fix it.
Treat AI like an intern who lacks context
Here is the line I keep coming back to. The reason AI hallucinates is that it just doesn't have enough context.
Think about what you would do with a new intern. You would not hand them your client calls and let them set prices in week one. You would give them your framework, sit them in on calls, show them the clients you priced well. Pricing well is something you trust to yourself, or to someone who has been in the firm for years.
An AI starts with none of that. So you give it the same things you would give the intern. A pricing framework, whether that is one you have built, one from your accounting alliance or a model like Hormozi's that suits where advisory pricing is heading. A discovery checklist, the four pages of questions you would ask on a call about entity type, industry, history and scope. And real meeting notes from calls you have actually had.
Documenting all of that is the highest-value work you can do right now. Everything you once used to train a person is what you will use to train the tools.
Build it as a project your whole team can use
The setup matters more than the interface. I build this as a Claude Project rather than a one-off chat, so the framework and files carry across every client and every team member can reach them. If you are a firm with a team, do this on a business or team plan. That keeps client data secure, and it means the work you put in is shared rather than trapped in one person's account.
On security, the rule is simple. Use a paid plan with training switched off before any client information goes near it. The free tier has its place. Keep confidential financials off it.
A few things make the output trustworthy rather than just tidy. Anchor it with an example of a client you priced well, never one you wish you had charged more for. Build in complexity adjustments for extra entities or states, and a cleanup and onboarding fee, the charge firms forget most often. Then feed it a real call transcript and ask it plainly what the discovery missed and how it would price the engagement.
When I did this live, it pulled the entity type, the transaction volume, the multi-state activity and the fact the books were months behind, then estimated $20,000 to $22,000 in revenue. The real value was elsewhere. The model told me what I still needed to ask.
Revenue is the easy half. Profit is the point.
A revenue figure on its own can flatter a bad client. So layer in your team's time. Survey how long the work takes, even as a range, and the tool produces a rough profit picture per client. That is where you find the engagements that look lucrative and quietly drain your capacity.
The same project keeps earning after onboarding. Drop in your ongoing meeting notes and ask where scope has crept. The client who started at $500 and now runs three entities will show up, and you will have the case to reprice them instead of absorbing the work.
From there you can generate the proposal in your own branding, and with Ignition and Anchor now connected to Claude, push it straight into your engagement tool.
Why this matters before 1 July
Here is the part I would sit with. I can build you a beautiful pricing app in an afternoon. If the framework underneath it is wrong, that app is more dangerous than no tool at all, because a clean interface makes people trust a wrong number more, not less. An app that confidently says $500 for a client who should be at $20,000 has just made your underpricing look official.
With Xero's pricing change and a new financial year landing, most firms are already reviewing fees. It is the right moment to fix the framework, not just the spreadsheet. Get the intelligence right underneath, and the tool on top will earn its place.
You can watch this session in full, along with all 17 from the AI in Practice Summit, at thefirm.media/summit.
