Key takeaways
- A fundable AI project fixes one manual process and plugs AI into how that part of the shop already runs.
- FedDev’s AI Adoption funding starts at $125,000 and covers up to 50% of eligible costs, so the project needs about $250,000 in eligible costs to reach the minimum.
- Four shapes fit most manufacturers: quoting, scheduling, quality inspection and predictive maintenance.
- Write it down on one page before you apply: the process, the data, the owner, the numbers you’ll measure and where your cash share comes from.
Deadlines
Project and eligible costs start before March 31, 2027: apply by November 24, 2026. All other projects: apply by January 19, 2027, 4:59 p.m. Eastern. Source: FedDev Ontario, when to apply.
New to the program? Read FedDev Ontario AI funding, explained first. This guide picks up where that one stops: shaping the project itself.
Start with one process
FedDev’s AI Adoption pillar funds the purchase and installation of AI solutions beyond commonly available AI tools. Its examples include AI readiness assessments, integrating solutions such as predictive analytics or robotics, data quality and governance, cybersecurity, and staff training (FedDev Ontario).
The practical test for a manufacturer: can you point to one process on your floor or in your office, describe how it runs today, and show where AI changes it? “We want to use AI” is a direction. “AI drafts every quote from the drawing and our job history” is a project.
Manufacturing and robotics is one of the sectors FedDev may give priority to. Priority helps. It doesn’t replace a clear project.
The size test
FedDev contributes $125,000 to $5 million per project, covering up to 50% of eligible costs. You fund at least the other half in cash (FedDev Ontario).
Work it backwards. At 50%, a $125,000 contribution needs $250,000 in eligible costs. A $40,000 software rollout doesn’t clear the floor, however useful it is. Projects that do clear it usually combine software, integration work, data cleanup, staff time and training around a single process.
Eligible costs must be reasonable, verifiable, incremental and directly tied to the project. Ongoing production costs, regularly scheduled capital spending, land, buildings, vehicles and the cost of preparing the application don’t count (application guide).
Four project shapes that fit manufacturers
Each one follows the same pattern: a manual process, the AI that changes it, the systems it connects to, and the numbers that show whether it worked.
1. Quoting from drawings and RFQs
Today: An estimator reads each drawing or RFQ, looks up similar past jobs and builds the quote by hand. Busy weeks mean slower quotes and missed bids.
What the AI does: Reads incoming RFQs and drawings, pulls comparable past jobs from your ERP, and drafts a quote with material, routing and machine time for the estimator to check and send.
What it plugs into: Your ERP or job history, your material price list, the inbox where RFQs arrive.
What you’d measure: Hours per quote, quote turnaround time, quotes sent per week, win rate.
2. Production scheduling and forecasting
Today: The schedule lives in a spreadsheet and one person’s head. Rush orders and machine downtime mean re-planning by hand.
What the AI does: Forecasts demand from order history and re-sequences jobs across machines when orders, materials or capacity change, with a planner approving the result.
What it plugs into: ERP or MRP, the shop-floor job board, purchasing.
What you’d measure: On-time delivery, changeover time, overtime hours, time spent re-planning.
3. Quality inspection
Today: Parts are checked by eye or by sampling at the end of the line. Defects are found late, after the material and machine time are spent.
What the AI does: Cameras and a trained vision model check parts on the line, flag defects as they happen and log the result against the job.
What it plugs into: Line cameras or new hardware, your quality records, the job or lot number.
What you’d measure: Scrap rate, rework hours, customer returns, inspection labour.
4. Predictive maintenance
Today: Machines are serviced on a calendar or when they break. A surprise breakdown stops a line and pushes every order behind it.
What the AI does: Reads machine data such as vibration, temperature and cycle counts, spots early signs of failure and books maintenance before the breakdown.
What it plugs into: Machine controllers or added sensors, your maintenance log, the production schedule.
What you’d measure: Unplanned downtime hours, maintenance cost per machine, missed ship dates.
A worked budget
Here’s the quoting project with made-up numbers, to show how a budget is built. Your costs will differ, and FedDev decides the actual contribution.
| Example quoting project | Amount | Type |
|---|---|---|
| Quoting software, licences and integration with your ERP | $110,000 | Capital |
| Server or cloud setup for the project | $30,000 | Capital |
| Outside expertise: data work, model setup, testing | $90,000 | Non-capital |
| Incremental labour: an estimator seconded to the project | $45,000 | Non-capital |
| Staff training | $15,000 | Non-capital |
| Total eligible costs | $290,000 | |
| Maximum FedDev contribution (up to 50%) | $145,000 | Repayable, interest-free |
| Your share, at least | $145,000 | Cash, not in-kind or future revenue |
Notice what’s missing: the cost of writing the application, ongoing subscriptions after the project ends, and normal production costs. None of those are eligible.
Total government help from all levels is capped at 50% of eligible capital costs and 75% of eligible non-capital costs, and SR&ED counts toward that cap (application guide).
What makes a project weaker
- It’s really a software subscription with nothing integrated.
- Nobody inside the business owns it.
- There’s no number to measure before and after.
- The data it needs doesn’t exist and the plan doesn’t fix that.
- The work started more than 12 months before the application date (how funding works).
- The staff count leans on contractors. FedDev counts full-time equivalents of at least 35 hours a week, and contractors don’t count (application guide).
The one-page scoping template
Fill this in before you talk to anyone about building it. If a line is hard to answer, that’s the part of the project to work on first.
- The process. Which one manual process are you fixing? Name it in one line.
- The cost of it today. Hours per week, people involved, and what goes wrong when it backs up.
- Your data. What records exist for this process today, where they live, and how clean they are.
- Systems it touches. ERP, MRP, inbox, machines, spreadsheets. List each one.
- The owner. Who on your team runs the project and signs off on changes.
- How you’ll measure it. Two or three numbers you track before and after.
- The budget. Capital and non-capital costs, listed separately. Total at least $250,000 if you want the $125,000 minimum.
- Your cash share. Where your half comes from: cash on hand, a bank line, retained earnings.
- Start date. Before March 31, 2027 means applying by November 24, 2026. Later starts have until January 19, 2027.
- Responsible AI and IP. Who owns the model and data, how errors get caught, and who checks the output.
Where SET fits
We help southern Ontario businesses plan and build AI projects: readiness assessments, the build plan and budget, and the systems themselves, such as quoting, follow-up, scheduling and reporting built into SET OS. Your business prepares and submits its own application, and funding decisions are made by FedDev Ontario. Our fees are fixed and never depend on approval. No outcome is guaranteed.
Start with the free AI funding checklist, or book a free strategy call and bring your one-page template.
This guide is general information, checked against FedDev Ontario’s pages on November 20, 2026. It isn’t legal, tax or financial advice, and FedDev can change its guidelines at any time. Confirm details with FedDev Ontario (1-866-593-5505) before you apply.