Why I Built sredy — and Everything Else That Happened Along the Way

The $30,000 invoice that started everything

In 2023, a founder friend — let's call him Daniel — showed me the invoice from his SR&ED consultant. $32,500 in contingency fees on a $108,000 claim. Daniel's company was a 6-person startup in Montreal. They had built a genuinely innovative ML pipeline for industrial quality control. The R&D was real. The documentation was decent. And they had paid 30% of their refund to someone who essentially formatted their existing notes into a CRA template.

Daniel told me: 'I felt like I had a choice between leaving money on the table or giving away a third of it. Neither felt right.'

I just wanted to see what AI could actually do

It started with ChatGPT. Pure curiosity — nothing more. I wasn't looking to build a company. I just wanted to see what would happen if I gave it a real tax document and asked it to explain the SR&ED eligibility rules in plain English. Then I asked it to draft a technical narrative from a set of Jira tickets. Then I asked it to interview me about my R&D work and turn my answers into a T661 section.

The outputs were rough. But they were also shockingly good for a first pass. I kept going.

I moved into Replit. Started actually building things — not just prompting, but writing real code with AI assistance, spinning up prototypes in hours instead of weeks. Then Claude. Then Codex. Each tool opened a new door. I'd close my laptop at midnight and realize I'd been at it for five hours without noticing.

What I kept coming back to was the SR&ED problem. Every time I pushed the AI harder — gave it more context, better prompts, real CRA guidance documents — the outputs got more defensible. More specific. More like what a good consultant would produce, but generated in minutes instead of weeks.

I realized: the bottleneck in SR&ED preparation isn't expertise. It's structure. It's having the right questions in the right order, connected to the right CRA criteria, with the output formatted correctly. That's a software problem. AI makes it solvable.

What I found talking to 200 Canadian founders

While I was building, I was also talking. Over 18 months I spoke with more than 200 Canadian tech founders about SR&ED. The same patterns came up over and over:

• 73% had never filed — mostly because they believed it was 'too complex' or 'not for companies our size'
• Of those who had filed, 68% used a consultant and paid 15–30% in contingency fees
• 41% of consultant users felt the value didn't justify the cost: 'they basically copy-pasted our Jira tickets into a Word doc'
• Sole proprietors were almost universally told SR&ED 'wasn't for them' by accountants
• The most common reason for not filing: 'I didn't know where to start'

The system was designed for aerospace companies with dedicated tax teams. Not for a 3-person startup in Kitchener, or a freelance developer in Halifax, or a healthtech founder in Vancouver spending 80% of her time on product.

What sredy actually does

sredy isn't a consultant replacement in the traditional sense. It's a guided system that helps you do it yourself — with the same quality output, at a fixed fee that doesn't take a percentage of your refund.

• Eligibility assessment: A structured questionnaire that identifies qualifying work based on CRA's two-part test. Most founders discover 20–40% more qualifying work than they initially thought.
• Technical narrative builder: Turns your existing documentation — Git commits, design docs, sprint notes — into CRA-ready narratives using AI-assisted prompting. Structured, specific, evidence-backed.
• Expenditure calculator: Calculates qualifying wages, materials, and overhead with the correct CRA allocation formulas. Connects to QuickBooks or Xero, or accepts manual entry.
• T661 form generation: Populates the required CRA forms automatically, with review checkpoints to catch the most common filing errors.
• Expert review option: A fixed-fee CPA review for complex claims — not a percentage of your refund.

What sredy doesn't do: it doesn't claim work that doesn't qualify. It doesn't inflate numbers. It doesn't promise refunds that aren't defensible. The goal is to help you capture the credits you already qualify for — nothing more.

The tool I accidentally built while building sredy

Here's something I didn't plan: while I was deep in building sredy, I kept running into the same operational problem. Founders were reaching out. People were interested. I needed to track conversations, follow up, remember who I'd spoken to, understand where someone was in their thinking. A basic sales and outreach workflow.

I could have just opened a spreadsheet. But by this point I was in full AI-builder mode. So instead I built a lightweight lead generation and CRM tool — purpose-built for sredy's outreach and pre-launch pipeline. It tracks prospects, conversation history, follow-up timing, qualification status, and stages each potential customer through a simple workflow.

The lead gen tool is now the system I'll use to manage sredy's go-to-market as we open to more founders. It's also a small proof of what building with AI feels like when you lean into it fully — a real working tool, built fast, because the technology has genuinely changed what one person can ship.

1.00 Second Baseball — built with my kids on a weekend

This one has nothing to do with SR&ED. But it has everything to do with why I keep building.

Growing up, my friends and I had a game we'd play with our digital watches. You'd start the timer, look away, and try to stop it at exactly 1.00 second. No cheating. No counting out loud. Just feel. It sounds simple. It isn't. Getting to exactly 1.00 — not 0.98, not 1.03 — takes a weird kind of focus that kids find completely addictive.

I told my kids about it one night. They were immediately obsessed. And then one of them said: 'Why isn't there an app for that?'

1.00 Second Baseball: you tap to swing. The game measures how close you got to exactly one second. Home run if you nail it. A double, a single, a strikeout — the outcomes scale with your precision. We added a baseball theme because it made the timing feel meaningful. A swing that's a hair too early, a fraction too late — that's the whole game.

We built it using the same AI-assisted tools I'd been using for sredy. The whole thing took a weekend. My kids tested every version, argued about the scoring, and lobbied hard for specific features. It's currently submitted to the Apple Store for review.

I don't know if anyone outside my family will ever download it. I don't care. Building it with my kids on a Saturday afternoon — watching them get excited about a problem that was solvable, and then solving it — was one of the best things that's come out of this whole AI obsession.

The mission: every Canadian founder files

The SR&ED program distributes roughly $3.1 billion annually. Canadian tech companies leave a significant portion of available credits unclaimed every year — not because they don't qualify, but because the system is too opaque, too expensive, and too intimidating.

sredy's mission is straightforward: make SR&ED accessible to every Canadian founder, regardless of company size, technical background, or budget. Fixed fee. Guided process. No percentages. No surprises. Just the credits you've already earned, in your bank account, without giving a third of them away.

If you've never filed SR&ED — start today. The documentation you need is already in your Git history, your design docs, your sprint notes. The only missing piece is the system that turns it into a claim. That's what we built.

And if you're also just curious about AI and what you can build with it — start that too. You might be surprised where it takes you.

Turn this guide into a claim package

SREDY.IO walks you through eligibility, project narratives, supporting costs, and evidence so your CPA has a cleaner file to review.

More guides

Ready to check if your work qualifies? Take the free eligibility assessment.