You Don't Need an AI Developer. Your Accountants Already Are One.

You Don't Need an AI Developer. Your Accountants Already Are One.

There is a version of the AI implementation problem that sounds unsolvable: you want to automate your firm's workflows, but you can't afford an AI developer, your firm doesn't have engineering staff, and every vendor quote you've received is priced for a company ten times your size.

Here's what CLA figured out: the limiting belief is the problem, not the resource gap.

CLA — the eighth-largest accounting firm in the US by revenue — built a 100-person AI tool-building program without hiring a single developer for that program. They did it by training their own accountants. In 2024, those accountant-built tools were used tens of thousands of times across the firm.

Their term for these accountant-builders: citizen developers.

Their insight: the people who best understand your workflows are the people already doing them. You don't need to hire someone who understands AI. You need to train the people who already understand your firm.

---

What CLA Did

CLA began experimenting informally with internal tool-building in 2022. The setup was minimal: staff members who expressed interest in building workflow improvements were matched with a mentor who knew the technical side. They worked on their own ideas or took projects from a shared backlog.

By 2024, the program was formalized. The citizen developer group had grown from 3-4 people to approximately 100 members — accountants, tax professionals, and operations staff from across the firm. These weren't IT staff. These were practitioners who had a workflow problem and decided to solve it.

The results: tens of thousands of tool uses in 2024 alone.

The tools they built ranged from individual task automations (send a Slack message when a client file is uploaded) to multi-step workflow systems (intake a new client, populate their record, create onboarding tasks, and send a welcome sequence — all without staff touching a keyboard). None of this required software engineering. It required patience, curiosity, and a few hours a week on a platform like Zapier or Microsoft Power Automate.

---

How the Citizen Developer Model Works

The CLA citizen developer model has three components:

1. Match — identify staff who want to build. This is the filter. Not everyone wants to do this. The citizen developer model doesn't require your whole firm. It requires finding 1-2 people who are already informally solving problems — the associate who built a better spreadsheet, the manager who rigged up an email filter. That's the person.

2. Mentor — pair them with someone who knows the platform. At CLA, the mentor had enough technical background to unblock the developer when they got stuck. At a smaller firm, the mentor might be you after watching a few hours of YouTube tutorials, or another citizen developer who started three months earlier.

3. Backlog — give them a list of problems worth solving. The biggest risk with citizen developer programs is that the first motivated person spends two weeks automating something nobody uses. Create a simple list of workflows that would save time if automated. Prioritize by frequency and friction. Give them clear targets.

That's the program. No engineering department required. No vendor contract. No dedicated budget line.

---

What This Means for a 10-30 Person Firm

The standard objection to the citizen developer model is that it only works at CLA's scale — 10,000+ staff, multiple regions, dedicated IT, formal training infrastructure.

This is backwards. Scale makes the citizen developer model harder, not easier. At CLA, 100 people were needed to have impact across a large organization. At a 20-person firm, two motivated people can build tools that affect every client engagement.

The math that matters isn't headcount — it's ratio. CLA needed 100 citizen developers out of thousands of employees. A 20-person firm needs 1-2 citizen developers. The odds that your current team includes one person who fits this profile are very high. The odds that you've given that person a structured path to do it are much lower.

The structural advantage of the smaller firm: tighter feedback loops. When a citizen developer at CLA builds a new tool, it might take weeks to see whether anyone uses it. When a citizen developer at a 20-person firm builds a client onboarding automation, the whole firm feels it within days.

---

Three Workflows Worth Starting With

If you're going to run a citizen developer program at a 10-30 person professional services firm, these are the three workflows with the fastest and most measurable ROI:

Client onboarding collection. Build a form — Google Form, Typeform, or JotForm — that collects all the information you need from a new client. Wire it to automatically create a client record in your practice management software, send a confirmation email, and create a task checklist for the engagement team. Every new client touches this. The time saved per client is small; multiplied across 50 new clients a year, it's significant.

Document request follow-up. Every accounting firm has the same problem: clients don't send documents on time. Build a simple automation that monitors your document collection status (from your practice management tool or even a shared spreadsheet) and sends reminder emails at 7 days, 3 days, and 1 day before a deadline. The firm wins because deadlines are met. The client wins because they're not surprised. The associate wins because they're not manually sending chase emails.

Deadline notification sequences. Tax deadlines, filing deadlines, extension deadlines — your clients forget them. Build an automated sequence that sends personalized deadline reminders 30, 14, and 3 days before each relevant date. This is a client service improvement with zero marginal cost after setup, and it's something no AI company can do better than you because it requires your client list and your deadlines.

All three can be built in a weekend using Make.com or Microsoft Power Automate at the $9-$29/month tier. Neither requires coding. Both require a motivated person with a workflow problem and a few hours to solve it.

---

The Talent Shortage Reframe

The accounting profession is facing a documented talent shortage. Open accounting roles have increased 8.5x in two years. Firms are competing for a smaller pool of credentialed professionals at increasing salary expectations.

The citizen developer model doesn't solve the talent shortage. It changes what the talent shortage means for your firm.

If you're waiting to hire your way into AI capability, you're waiting for a problem that won't resolve. The competition for AI-literate accounting staff will intensify through 2026 and 2027. The firms that come out ahead are the ones that are building AI capability into the people they already have.

CLA's citizen developer program is not a technology program. It's a talent development program that uses technology as the output. The accountants who went through it didn't become developers. They became more capable accountants — people who could see a workflow problem and solve it, not just do it.

That's the skill that compounds.

---

One Action This Week

Identify one repetitive workflow in your firm — something that takes 20-30 minutes every time it happens and happens at least weekly. Write down exactly what happens in that workflow, step by step.

Then find one person on your team who would enjoy automating it. Show them the Make.com free trial. Give them the workflow doc. Ask them to spend two hours figuring out if it can be automated.

That's the CLA model at its smallest possible scale. One workflow. One person. Two hours.

If they come back with a working automation, you've found your first citizen developer.

If they come back with questions, you've found your first candidate for a short training investment.

If they come back with nothing, you've learned something about that person that's also useful to know.

The program starts with a single conversation. The 100-person program CLA has today started with three or four people who said yes to a similar one.