QuarterlyFebruary 1, 20257 min read

HeitechSoft Quarterly: Winter 2025

November 2024 to January 2025

This quarter, AI moved from our lab into real work, at a client’s office and in our own. In our first Efficiency Edge assessment, for a family-run HVAC contractor, AI drafted the report from the interviews we recorded on site. It showed one dispatcher spending about four hours a day copying data from one spreadsheet into another, work a single button could replace. After the beta, Efficiency Edge opened to clients in January. Our R&D team compared Google Gemini, Microsoft Copilot Studio and Azure AI Foundry hands-on, and started building our own assistant that answers from our project files. We also ran more of our own business on AI, from Microsoft 365 Copilot and AI meeting notes to a website agent that captured its first lead and an AI review of our new office lease. We spoke to about 15 CEOs at the Greater KW Chamber on the risks and rewards of AI, and recorded a podcast on our pivot to AI. We also hired our first employee, an analyst chosen for AI skills, with AI video interviews

Introduction

This quarter AI left the lab and went to work, at a client's office and in our own. In our first Efficiency Edge assessment, for a family-run HVAC contractor, AI drafted the report from the interviews we recorded on site. One finding was almost too ordinary to notice: a dispatcher spending about four hours a day copying data from one spreadsheet into another, work a single button could replace. In my years as a consultant I kept running into businesses with a job like that hiding in plain sight. With the beta behind us, Efficiency Edge is now open to clients.

— Tom Heiber, CEO

Four hours a day, two spreadsheets, one button: our first Efficiency Edge assessment

Efficiency Edge is built for established, family-run businesses with roughly 20 to 70 employees, whose systems grew up over time rather than being designed. Our first assessment, in the beta, was with a family-run HVAC contractor. From late November to early January we sat with the people who do the work, one department at a time, and recorded the interviews.

AI transcribed those interviews and drafted the assessment report, and a person reviewed it before the client saw it. The owner gets findings drawn from how the business actually runs, and AI handles the slow work of turning hours of interviews into a written report.

One finding stood out. A dispatcher spent about four hours a day copying data from a Google spreadsheet into a Microsoft spreadsheet, work a single button could replace. The biggest savings often sit in plain manual work between two systems, not in anything exotic.

With the beta finished, we opened Efficiency Edge to clients in January. The service itself, from its process and description to its name and web page, was designed with AI.

Lab notes

Google or Microsoft for an AI assistant that answers from your own documents? We tried both

Clients keep asking for AI assistants that answer from their own documents, a technique known as retrieval-augmented generation. From December into January our R&D team tried Google Gemini, Microsoft Copilot Studio and Azure AI Foundry hands-on, then wrote up what we found. For firms already on Microsoft 365 we lean to Copilot Studio, which is mostly no-code, and to Azure AI Foundry when the assistant needs to complete tasks, not just answer. Most firms should start from the suite they already pay for.

We run our own business on AI first, from meeting notes to our office lease

We use AI for our own work first, and judge paid tools on what they return. This quarter we added Microsoft 365 Copilot. AI takes the notes and summaries for client, hiring and internal meetings, and we draft proposals with ChatGPT. Our website's AI agent answers visitors' questions and captured its first sales lead in January. When we negotiated our new office lease, AI reviewed every revision. That is how we learn what each tool costs and where it falls short before we recommend it to a client.

Wins

About 15 CEOs, one morning, and the risks and rewards of AI

The Greater Kitchener Waterloo Chamber of Commerce asked us to speak to its CEO Peer 2 Peer groups, combined for one session. On December 4, in "AI – Risk and Rewards", about 15 CEOs heard which AI tools small and mid-sized businesses can use now, what is coming, and how a business can benefit. The same day two business owners asked for help: one wanted to see where AI fits in copywriting, proposals, CRM automation, social media and reporting, and the other had tried Microsoft Copilot Studio and found it overwhelming.

Our first employee is an analyst hired for AI skills, and AI helped us find them

Demand for help putting AI to work kept growing, so we hired our first employee: a technical business analyst chosen for AI skills, who joins us in February. We put AI in the job title and asked for familiarity with AI tools, and early interviews showed how rare hands-on AI experience still is. AI video interviews then screened 61 of the 75 applicants on behavioural and soft skills, so a small team could consider a large pool. The role also adds capacity for Efficiency Edge assessments.

A podcast on our pivot to AI, and a school that asked for an AI expert

In January we recorded a podcast interview on how and why HeitechSoft pivoted to AI. A school-based careers forum asked for an AI expert, and we accepted an invitation to speak to more than 200 Grade 7 and 8 students in May about the AI careers likely to exist in the next five to seven years. Those shaping the next generation of workers are looking for practical AI voices, and we are glad to be one.

By the numbers

6
AI talks, panels and expert sessions in 2024. Each talk shows us which questions owners are actually asking, and those questions shape what we test next.
3
AI platforms we compared hands-on for business AI assistants. Hands-on comparisons show which route fits a firm's existing software before any money is spent on a build.
4
hours a day of copy-and-paste work we found in one client role. Manual copying between two systems is often the cheapest place to find time back for your team.

AI for mid-market

Ontario sets a 2026 start for AI disclosure in job postings

Ontario has filed the regulation under the Employment Standards Act that starts new job posting rules on January 1, 2026, including a requirement to say so in public postings when employers use AI to screen, assess or select applicants. The definition of AI is broad, and the rules do not apply to employers with fewer than 25 employees. If you have 25 or more staff, check now whether the screening and ranking built into your job boards or applicant tracking system counts, so you are ready before January 2026.

Source: Littler

Microsoft 365 Copilot Chat: free AI chat for business users, with agents billed by use

Microsoft has made Copilot Chat available to Microsoft 365 commercial customers at no extra cost, with enterprise data protection, so staff can use an approved AI chat instead of pasting work into personal accounts. Agents that staff can use inside the chat are priced on a metered basis with IT in control, which lets you test a use case with a few people before buying the full Copilot licence. Decide who may create agents and watch the bill, because usage-based costs are hard to predict.

Source: Microsoft 365 Blog

Google Workspace now includes Gemini AI in Business and Enterprise plans, with a small price rise

Google has folded Gemini into its Workspace Business and Enterprise plans rather than selling it as an add-on, so AI drafting, meeting notes and document summaries now come with the subscription. In Google's example, Business Standard with Gemini costs US$14 per user per month, $2 more than without it, and existing customers see the change from March 17, 2025, or later at renewal on annual plans. If you run on Google Workspace, decide which features to switch on and give staff a short guide before they use them on client work.

Source: Google Workspace Blog

Canada's proposed AI law dies as Parliament is prorogued

Parliament was prorogued on January 6, 2025, until March 24, which ended all bills in progress, including Bill C-27 and its proposed Artificial Intelligence and Data Act. It is now uncertain when, and in what form, Canada will have federal AI regulation. There is no federal AI law to wait for, and the privacy law that applies today, PIPEDA, still governs the customer and employee information staff put into AI tools, so a short policy on approved tools and permitted data is the practical step.

Source: DLA Piper

Microsoft adds ready-made Copilot agents for SharePoint, meetings and projects

At Ignite, Microsoft made agents in SharePoint generally available, so users can create assistants scoped to specific content, and put a meeting-notes agent for Teams and a project-management agent for Planner into public preview. Several of these are still previews, so treat them as things to try rather than rely on. An agent over a SharePoint folder of policies or product sheets is the easiest place to start, but its answers are only as good as the documents it can see, so tidy those folders and check who has access first.

Source: Microsoft 365 Blog

If any of this touches a process 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.