QuarterlyAugust 1, 20268 min read

HeitechSoft Quarterly: Summer 2026

May to July 2026

In 30 seconds

This quarter, we won the Guelph Chamber of Commerce's Innovation Award, which we see as recognition of a practical approach to AI: since ChatGPT arrived in late 2022, we have tested AI on our own operations before recommending it.
Our lab notes cover how we choose AI models by cost and show clients what their AI costs to run, how we plan software before AI coding agents write it, and why we check AI meeting notes against the recording.
Our wins include a buyer-matching platform for an M&A advisory firm, delivered in about five weeks, and an AI assistant for a pressure-vessel compliance start-up that answers only from records each user is allowed to see.
This issue covers five industry developments, including Statistics Canada's finding that one in five businesses now uses AI and the Bill C-36 privacy proposals.

Introduction

This quarter was about a way of working we have held to since ChatGPT arrived in late 2022: test AI on our own business first, then recommend only what earns its place. What surprised me most was hearing our name called as the winner of the Guelph Chamber of Commerce's Innovation Award, at the Guelph Awards of Excellence on June 18. It is a welcome recognition of plain, hype-free AI work. This issue shows what that looks like in our lab, in our client work and across the wider market.

— Tom Heiber, CEO

A practical approach to AI, recognised by the Guelph Chamber of Commerce

On March 31 the Guelph Chamber of Commerce's Awards Committee told us we were one of three finalists for its Innovation Award, chosen from a high number of nominations. On June 18, at the Guelph Awards of Excellence, we won it.

For us, the award reflects a decision we made when ChatGPT arrived in late 2022. We believed AI would change how businesses operate, so we pivoted to become an AI-first company. Before recommending AI to anyone, we tested it across our own operations, research, documentation, content and software development.

AI does not improve a business just because it signs up for a new tool or adds a chatbot to its website. In growing businesses of 20 to 50 people, we see a mix of CRMs, spreadsheets, accounting systems, shared drives, project tools and email threads. Together they create duplicate data entry, confusion and cost. Most of those businesses need fewer disconnected tools, cleaner workflows and smarter systems, not another AI subscription on top.

For an owner, the lesson is to fix how work moves first: remove duplication and disconnected tools, then let AI make the cleaner system faster.

Lab notes

Newer AI models cost more, so we choose by task and show the bill

When Anthropic released its newest top model in June, we tested it on real projects. It costs about twice as much as the previous top model, so we use it where it earns that cost and cheaper models for the rest. Inside client systems, a low-cost Google Gemini Flash model handles most routine work. We also price the AI a system uses separately from building it, and built a portal that tracks each client's AI costs by provider, with a budget forecast. Ask any provider which model your system uses, and why.

AI writes code fast, so the planning is the real work

Most of our software is now written by AI coding agents, and the quality comes from the structure around them. On our two largest builds this quarter, each project started with a written specification, split into phases and small tickets, each checked against about 20 quality gates. One agent builds each ticket, another reviews it, and a developer reviews the result. As the developer who runs our pipeline puts it: "Most time is spent on the planning phase. Once I have that, I let my agents go on autopilot and build end to end." When you commission software, ask how AI-written code is checked before it reaches you.

AI meeting notes save time, and we still check them against the recording

AI note-takers sit in on most of our meetings and produce summaries, which saves real time. We have learned to check them. On a call that switched between languages, neither of the two note-takers we tested detected the change. AI summaries also get facts wrong: one presented a one-time backfill as a daily rate, and another put a client's decision deadline a month later than the date given. Treat AI notes as a first draft: before acting on a figure or date from a summary, check it against the transcript.

Wins

Buyer matching for an M&A advisory firm, delivered in about five weeks

An M&A advisory firm matched businesses for sale with likely buyers using spreadsheets, inboxes and manual research. It signed with us on May 4, received a first release on May 13, and the agreed scope was complete on June 9. AI reads company profiles and buyer information from documents and proposes the data, but an advisor reviews each field before anything is saved. Matching follows the firm's own hard requirements and a weighted fit score, and each buyer comes with a plain-English reason, so an advisor can check it before making the call.

A pressure-vessel compliance start-up adds an AI assistant that respects permissions

A start-up's platform keeps the compliance paperwork behind every pressure vessel in one place. Its founder accepted the minimum viable product on May 8, then signed a third phase, and on May 29 we released an AI assistant. It answers plain-language questions across vessels, documents, sign-offs and inspections, such as which vessels are missing a test report. It works under each user's own permissions, so it never shows a record that person couldn't open, and every answer links to the records behind it. In May the founder wrote that we "delivered everything on time and on budget."

By the numbers

98%
of the code changes in our own software this quarter (122 of 124) were made with an AI coding agent.
Almost every change to our own software now goes through an AI coding agent, so our advice on these tools comes from daily use.
580
pull requests on our two largest client builds, made through our AI coding-agent pipeline.
Small, reviewable changes are what keep large AI-assisted builds easy to check.
13
releases of a start-up's compliance platform in three months.
Frequent releases mean changes reach working software quickly, instead of waiting for one large launch.

AI for mid-market

Ottawa's new AI strategy targets 60% business adoption by 2034

The federal government released AI for All, its national AI strategy, which aims to raise business AI adoption from about 12% to 60% by 2034. For a firm your size, most of the money will flow through existing channels such as BDC financing and FedDev Ontario, and the strategy gives no intake dates for the new funds, so ask both what is open before you budget a project. By the strategy's own count only about 8% of SMEs have adopted AI, so a firm starting now with one measurable use case is not behind.

Source: Innovation, Science and Economic Development Canada

Bill C-36 proposes new privacy rules for businesses, including how AI decisions are explained

The federal government tabled Bill C-36, the Protecting Privacy and Consumer Data Act, which would modernise private-sector privacy law and be enforced by a new regulator able to issue binding orders, with penalties of up to $10 million or 3% of global revenue. Businesses would need meaningful consent with clearer explanations, more transparency about automated decision systems including AI, and a privacy risk assessment before sending personal information outside Canada. The bill is only at first reading with no coming-into-force date, so PIPEDA still applies, but if AI helps you screen applicants, approve customers or set prices, start writing down what it does and where the data goes.

Source: Innovation, Science and Economic Development Canada

Statistics Canada: one in five businesses now uses AI, and privacy worries are the top barrier

Statistics Canada reported that 19.2% of businesses used AI to produce goods or deliver services in the 12 months before its April to May 2026 survey, up from 12.2% a year earlier, and that cybersecurity or privacy concerns were the top barrier. Adoption has roughly tripled in two years, though it is still under 10% in construction and wholesale trade, so how far you lag depends on your sector. If privacy concerns hold your firm back, as they do for more than one in five firms your size, a short written rule on approved tools and what data may go into them addresses that directly.

Source: Statistics Canada

Microsoft makes its Copilot bundles for small business permanent

Microsoft told partners that from July 1, 2026, Microsoft 365 Business Standard with Copilot and Business Premium with Copilot become permanent plans, at list prices of US$23.50 and US$32 per user per month. Against the base plans at US$14 and US$22 a user per month, the bundles add Copilot for about US$10 a user, roughly half the list price of the separate add-on. These are US-dollar list prices, so ask your Microsoft partner for Canadian pricing and compare a bundle with your current plan at your next renewal.

Source: Microsoft Partner Center

Statistics Canada: more than a third of workers now use generative AI at work

In a March 2026 Labour Force Survey supplement, 35.9% of workers aged 15 to 69 said they had used generative AI tools at work in the past 12 months, from 65.6% in professional, scientific and technical services to 16.3% in accommodation and food services. Most users apply these tools to some tasks rather than reshaping their jobs. In an office-based business, expect that a third or more of your staff already use these tools, whether or not you have approved them, which makes now a good time for a one-page policy on approved tools and customer data.

Source: Statistics Canada

Written by

Tom Heiber, Founder & CEO, HeitechSoft

Tom founded HeitechSoft in 2004. Since 2022 he has run it as a working “AI company experiment”: applying AI across the business and sharing what actually holds up in practice.

LinkedIn About HeitechSoft Talk to us

If you want to see where AI fits in your business, book a 20-minute conversation at https://heitechsoft.com/contact.

Get the next issue by email

Each quarter we share what we built, what we learned and the AI news that matters to mid-sized businesses.