How I use Ignition’s MCP to automate jobs I used to dread
Copied to clipboard.
A few weeks ago, I asked Claude to roll one of our client proposals forward, apply a price increase, and draft the client email.
It did the whole thing in a couple of minutes. I didn’t open Ignition once.
But the speed wasn’t the part that stuck with me.
The bit that made me sit up was that it left one line alone: an SMSF audit fee we simply pass through. It saw that the line shouldn’t get the increase, treated it differently, and moved on.
That was the moment MCP stopped feeling like another AI acronym and started feeling useful.
MCP in plain English
If you’ve heard the term MCP lately and quietly nodded along, you’re not behind.
MCP stands for Model Context Protocol. In plain English, it’s a standard way for an AI assistant, like Claude or ChatGPT, to connect to the tools your firm already uses and work with them on your behalf. In Ignition’s case, that means your assistant can look things up in your account and, with the right approvals, help take action in everyday language instead of you clicking through the app yourself.
Think of it as plumbing.
You don’t see it, but it connects the assistant to your systems. The AI is the tap. You’re still the one deciding what comes out.
That distinction matters. This isn’t “let the robot run the firm.” It’s more like giving a capable assistant access to the right filing cabinet, with you still checking the work before it goes anywhere important.
Why this matters for a firm
For most of us, AI started as a smarter chatbot.
Write this email. Summarise that note. Tidy up this paragraph.
Useful, yes. But you still had to pick up the tool every time.
The bigger shift is AI that can help with repeatable operational work: reviewing proposals, updating services in bulk, supporting onboarding, checking renewals, and turning messy firm information into a clear next step.
That’s where this starts to get interesting.
Not because it’s flashy. Because it fits the work we actually do.
What I’ve actually run
1. Repricing a proposal
This is the one I opened with.
Claude rolled a proposal forward, applied the price increase, drafted the client email, and left a pass-through SMSF audit fee untouched.
The lesson wasn’t “AI is fast.” We already know that.
The lesson was that it didn’t blindly find and replace a number. It looked at the job, treated different lines differently, and gave me something I could review instead of something I had to rebuild.
2. Tracking renewals
Renewals have been the big one this year.
Instead of babysitting a spreadsheet, I ask Claude to pull our whole book of proposals into one view:
Who’s due to renew?
Who’s still on last year’s price?
Who has signed?
Who hasn’t?
Where are we quietly leaving money on the table?
That used to be a job I dreaded. Now it’s a review.
And that’s the pattern I keep coming back to. The work doesn’t disappear, but the painful part gets smaller.
3. Refreshing proposal terms in bulk
I also refreshed the terms on every active proposal in the firm. More than 300 of them.
Normally, that kind of job is days of clicking, checking, second-guessing, and hoping you didn’t miss one.
This was one run, with review points along the way.
I’m not saying you should blindly push changes across a few hundred client agreements. You shouldn’t. But being able to prepare, check, and manage that work at scale changes the job completely.
4. Charging for ad-hoc advice we used to write off
This one surprised me.
We’ve all had those messy advice jobs that start as a quick question, turn into a string of emails, then a chat thread, then a couple of voice notes, then somehow become “just part of the relationship.”
Now I can ask Claude to read the engagement context, emails, chats, and even the length of the voice notes, then pull together a timesheet for me to check.
Once I’ve reviewed it, Claude can draft the proposal in Ignition from a single prompt.
Different job, same pattern: it works with the real, messy way a firm actually runs.
See how Ignition MCP works
Want to try this inside Ignition? Start with the official setup guide and connect your AI assistant to your Ignition account.
What to know before you try it
A few honest pointers.
Start read-only.
Ask it to read and summarise your data before you let it change anything. You’ll learn quickly where it’s helpful and where your prompts need tightening.
Keep a human on anything that touches money or a client.
AI can inform the decision. It shouldn’t own the decision. Pricing, billing, renewals, and client communication still need your judgement.
Test on two or three before you run the herd.
I once hit my Claude usage limit while running a few hundred proposal updates in one hit. Big jobs need pacing.
Only connect tools you trust.
Your assistant works through the permissions you give it. That’s powerful, but it also means you need to be deliberate.
This is the boring part of AI adoption. It’s also the part that keeps you out of trouble.
What I’ve learned
The tech was never the hard part.
The hard part was working out where AI actually fits, what’s safe to connect it to, and how to get real value instead of another experiment that fizzles out after a week.
The firms that win with this won’t be the ones with the fanciest tools.
They’ll be the ones that pick one real, repetitive, painful job, prove it out, then move to the next one.
The first one is hard.
The tenth is a Monday afternoon.
Where Ignition fits
For us, a lot of this runs through Ignition because that’s where our proposals, clients, billing, and renewals live.
Ignition’s MCP is currently listed as an Early Access Labs feature, available to all Ignition customers, and it lets AI assistants like Claude and ChatGPT/Codex interact directly with your Ignition account so you can ask questions about your data and automate administrative tasks.
You turn it on under Settings → Labs, connect your assistant, and start small.
Don’t begin with “update everything.”
Begin with “show me the renewals due this month” or “summarise which proposals are still on last year’s pricing.”
That’s where the confidence comes from.
Not ready to connect AI to your systems yet?
Start with AI Price Insights in Ignition. It gives you data-backed pricing guidance inside the proposal builder and service library, including pricing benchmarks, suggested price points, and reasoning behind the recommendation.
My advice: don’t start with AI. Start with the job.
The trap with AI is asking, “What can this tool do?”
That’s how you end up with clever demos and no change in the firm.
The better question is: “What job do we do over and over that is annoying, important, and easy to check?”
For us, that’s been renewals, pricing, proposal terms, and ad-hoc advice.
For your firm, it might be onboarding. Or scope checks. Or chasing old proposals. Or finding clients who are still on legacy pricing.
Pick one.
Make it safe.
Run it twice.
Then decide if it earns a place in the firm.
That’s the difference between playing with AI and actually using it.
See it in action
I’m joining Inbal Rodnay on 20 July for a practical session on MCP and agents for accounting firms.
I’ll walk through what we’re actually running inside the firm, where I’m still cautious, and how I’d suggest getting started without turning your practice into an AI science experiment.
Come along.
Want to see MCP in action?
Join Ben Walker and Inbal Rodnay for a practical session on MCP and AI agents for accounting firms.
