QuarterlyOctober 30, 20258 min read

HeitechSoft Quarterly: Fall 2025

August to October 2025

This quarter, we moved from telling owners what AI can do to showing them. Our lead story is the public launch of AI Explained, with a website drafted live at a Greater KW Chamber of Commerce lunch and sessions tailored to each owner's own work. In the lab, we used AI research for personal outreach with a person checking every email, had an AI agent build a restaurant group's inventory prototype in under three hours, turned a script into an AI-made ad video in an afternoon, and built a board of AI advisers. Our wins include the Chamber's Expert Exchange, a Waterloo hotel group's management team choosing its first AI pilot, and an article on Canada's AI paradox. Our AI for mid-market section covers five industry developments, including Statistics Canada's survey of AI plans and OpenAI's company knowledge feature for ChatGPT Business.

Introduction

This quarter was about proof. We launched AI Explained publicly, showed AI doing real work in front of owners, and used AI to research our own outreach, with a person checking every email before it went out.

The moment that surprised me most: a multi-location restaurant group replied to an email our AI had researched and drafted, and five days after our first call it had a working prototype of the inventory tool it wanted. That speed is new, and it is why I now show owners AI at work instead of describing it.

— Tom Heiber, CEO

We stopped telling owners what AI can do and started showing them

In early September we launched AI Explained publicly, our one-on-one AI coaching for business owners and leaders. We printed 3,000 flyers and made promo videos from AI-generated clips.

On October 2 we gave a live AI crash course at a Greater KW Chamber of Commerce networking lunch for about ten owners. AI deep research built a brand document on one attendee's firm, and an AI coding agent drafted a first version of the attendee's website in about five minutes while the room watched. We said the caveats out loud: it was a first draft, and AI can still leave security holes. The attendee called the new website their win of the week.

Consultations since the launch shaped the offer. A downtown business association's marketing lead told us they had done plenty of AI training and walked away from every session knowing no more than before. A theatre's director wanted AI to complement the theatre's people, not replace them.

So each session is tailored to the owner's own work and shows AI doing a job they recognise. A short session on your own work, using tools you already pay for, is a low-risk way to find a first practical use.

Lab notes

AI researches each business before we write, and a person checks every email

In September we built a pipeline in which AI reads a local business's website, researches the business and its industry, and drafts a personal invitation to an AI Explained consultation. More than 400 went out from September 22 to October 17. An acquaintance found one impersonal, and the AI could not know who we already knew, so since early October a person checks every email and every claim before it goes. The emails opened conversations with a restaurant group, a theatre and a business association. AI can do the research; deciding what to send stays with a person.

A restaurant group's inventory prototype, built by an AI agent in under three hours

A multi-location restaurant group replied to one of those emails. It was piecing together US-based subscription tools with limited features in Canada. After our first call we wrote down what an inventory tool had to do and gave it to two AI coding agents. Under three hours of agent work produced a working web app with stock counts, purchase orders, waste tracking and reorder suggestions, which the operations manager called very cool and intriguing. It is a prototype, and a software engineer still checks the code, but a tool built around your own jobs can now be tested in days.

A script to an ad video in an afternoon, and what AI video still cannot do

A large US marketing agency asked what AI video could do for an online product listing. We sent a rough cut the same afternoon the script arrived, and a second cut with the logo two days later. We were open about the limits: eight-second clips, no 4K output, and glitches that mean some clips must be generated again. We suggested a second AI to check each clip for consistency before a person reviews it. In September we made it a service with a first draft within 24 business hours.

We gave ourselves a board of AI advisers

In August we set up custom Google Gemini assistants as a CEO, marketing and CTO adviser. All three work from one shared knowledge base about our company, services and goals. We use them to test ideas and decisions before we act, at any hour. Any owner can build a small sounding board like this from documents they already have. Its advice is only as good as what you tell it about your business, and the decision stays with you.

Wins

Back at the Chamber's Expert Exchange, a year on

On October 30 we joined the tech edition of the Greater Kitchener Waterloo Chamber of Commerce's Expert Exchange at Communitech, a year after our first. We hosted the productivity-tools table, where AI was a big part of the conversation. The next day the Chamber wrote that guests really enjoyed the session, and that so many said they walked away with new ideas and insights. Our takeaway for owners: a short conversation about one tool for one job is often all you need to start.

A Waterloo hotel group's leaders, after one manager's crash course

A manager at a Waterloo hotel group took our AI crash course this summer. In September the general manager, financial controller, and sales and food-and-beverage leads sat down with us for a session of their own. The sales lead summed up the problem: "we're doing double the job using AI right now because we're not using it effectively." The team chose a first pilot: an AI-built social media calendar for its 2026 food and beverage events. The gain comes when AI sits inside the workflow, not beside it.

Canada's AI paradox: world-class research, a business reality check

On September 3 Tom published an article on Medium that sets Canada's standing as a global centre of AI research against a KPMG study ranking Canada 44th of 47 countries for AI literacy. The article calls the gap 'a five-alarm fire for our national competitiveness' and argues it is one of translation, not technology. The same gap exists inside most businesses: the tools are already there, and what is missing is someone to translate them into your own work.

By the numbers

400+
personal emails drafted from AI research on each business, sent September 22 to October 17. Shows how AI research can prepare personal outreach for hundreds of businesses, while a person still decides what goes out.
2
AI sessions at Greater KW Chamber of Commerce events this fall, a live crash course and the Expert Exchange. Hands-on sessions on real work let an owner see what AI can do before deciding whether to pursue it.
179
applications for our five AI-first internship and engineering roles since mid-August. A sign of how many people want to build AI skills, which matters for any business planning its own hiring.

AI for mid-market

StatCan: 15% of Canadian firms with 20 to 99 staff plan to use AI in the next year

Statistics Canada reports planned AI use at 15.0% for firms with 20 to 99 employees, while 66.7% of all businesses have no plans. For a firm of 20 to 200 people, a modest pilot on one repetitive office task, such as quotes or customer email, can put you ahead of competitors who have not started.

Source: Statistics Canada

Copilot Chat now built into Word, Excel, PowerPoint and Outlook at no extra cost

Microsoft 365 customers can now use Copilot Chat inside Word, Excel, PowerPoint, Outlook and OneNote without paying extra, while the paid Microsoft 365 Copilot add-on still draws on your organisation's work data. Most of your staff may already have an AI assistant in the apps you license, so set a short rule on what data can go in and who checks the output before anyone relies on it.

Source: Computerworld

Ottawa launches AI Strategy Task Force and 30-day public consultation on next national AI strategy

The public consultation on Canada's next AI strategy ran from October 1 to October 31, and adoption across industry is one of its eight themes. For a business of 20 to 200 people, keep a short note of your main barriers (cost, skills, data) and ask your industry association or chamber what it submitted, so you are ready for any follow-up.

Source: Innovation, Science and Economic Development Canada

MIT report: only about 5% of AI pilots deliver rapid revenue gains

MIT's NANDA initiative, based on 150 leader interviews, a survey of 350 employees and 300 public AI deployments, found that buying tools from outside vendors or building partnerships succeeded about 67% of the time, while internal builds succeeded only one-third as often. For a 20 to 200 person firm, that points to buying proven tools and starting with back-office work such as invoicing or reporting, where results are easier to measure.

Source: Fortune

ChatGPT can now search your company's Slack, SharePoint and Google Drive

OpenAI's company knowledge feature lets ChatGPT Business users pull answers from connected work apps, with citations to the original sources, and ChatGPT only sees what each user is already authorised to view. That makes it more useful, but anything in an over-shared drive also becomes easy to find, so have your admin review which connectors are on and tighten broad sharing before staff start using it.

Source: VentureBeat

If you want to see what AI could do with your own work, 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.