> Site index: [llms.txt](https://ortelian.com/llms.txt) with all pages and descriptions. # How do you get staff who aren't comfortable with software to adopt AI agents? Source: https://ortelian.com/notes/how-to-get-staff-to-adopt-ai-agents/ Markdown: https://ortelian.com/notes/how-to-get-staff-to-adopt-ai-agents.md Author: Laurens Nys Published: 2026-11-04 Staff adopt AI agents when the agent removes work instead of adding a tool: build it with them, meet them in tools they know, start with the work they dislike. Staff who aren't comfortable with software adopt an AI agent when it takes work off their plate without asking them to learn a new tool. Build it with the people who do the work, put it in the tools they already use, start with the part of the job they like least, show them what it did and why, and give them one person to ask when it is wrong. Training matters less than any of these. ## Why adoption fails in teams like this We hear the same worry on calls with buyers. One described older staff and volunteers who leave if something feels like extra work. Another said that the moment it gets too technical, companies say yes and then don't do it. A third pointed at the time it takes people to get used to any new tool. Research on why people take up new technology says much the same. The Technology Acceptance Model found that two things predict use: whether people think it helps them do the job, and whether they think it is easy to use ([Davis, 1989](https://doi.org/10.2307/249008)). A later model that combined eight earlier ones added that people use what their colleagues expect them to use and what they have support for ([Venkatesh and others, 2003](https://doi.org/10.2307/30036540)). None of these is about how clever the AI is. They are about the extra work it asks for. So the usual rollout, a new app, a training session and a login, fails exactly the people it is meant for. It adds a screen, a password and a new way of working on top of the old one. ## Five steps that make AI agents easy to adopt 1. **Build it with the people who do the work.** Watch the job, ask them where it gets stuck, and test early versions on their real cases. People trust a tool they helped shape, and they spot the exceptions nobody wrote down. 2. **Meet them in the tools they already use.** The agent sends a draft to their inbox or chat, or leaves it in the system they already work in. Nobody has to open a new app. How that looks day to day is in [how employees use AI agents day to day](https://ortelian.com/notes/how-employees-use-ai-agents-day-to-day/) . 3. **Start with the work they like least.** Retyping orders, chasing missing fields, copying data between systems. If the first job removes a chore, the agent earns goodwill. If it changes the part of the job people are proud of, it earns resistance. 4. **Show what it did and why.** Every draft carries its sources and what is unusual about this case, so checking it takes seconds. A person who can see why the agent did something is more likely to rely on it. See [how to trust AI output without checking every line](https://ortelian.com/notes/how-to-trust-ai-output-without-checking-every-line/) . 5. **Give them one person to ask.** Someone inside the team who knows what the agent does and takes the call when it is wrong. A fix that comes the same week tells people their feedback counts. Who that person is, and what they own, is in [who owns the AI platform after launch](https://ortelian.com/notes/who-owns-the-ai-platform-after-launch/) . ## What to ask people for, and what not to | Ask them to | Don't ask them to | | --- | --- | | Approve, edit or reject a draft in their inbox or chat | Learn a new app or log in somewhere new | | Say when the agent is wrong, in their own words | Write prompts or fill in a feedback form | | Show you how they handle the messy cases | Follow a new process designed without them | | Decide the exceptions the agent can't settle | Recheck every line the agent produced | ## How to tell if it is working Count the human effort, not the logins. Track how many drafts people approve without changes, how often they correct the agent, and how long the review takes, against how long the work took before. If people quietly go back to doing the job by hand, the agent added work somewhere. Ask them where, and fix that before you add the next job. More on measuring this in [how to measure AI agent ROI](https://ortelian.com/notes/how-to-measure-ai-agent-roi/). ## How this works with Ortelian Our forward deployed team spends time on site observing the work and talking it through with the people who do it, and tests what we build with them. Agents work in the tools your team already uses, and actions that leave the company, such as sending an email or an invoice, wait for a person. Every action is recorded, with the sources behind it. We track quality, cost, time and human effort on real examples of your work. Workshops during the build prepare your team to own the work. After launch, your team runs it or we keep running it for you; either way, we host and operate the platform. For how an engagement runs, see [how we work](https://ortelian.com/how-we-work/). ## Questions people ask ### Do staff need AI training before an agent goes live? Not much, if the agent works in tools they know. They need to know what it does, what it will ask them and who to tell when it is wrong. Show them on their own cases, not with general AI training. ### What if some people refuse to use it? Find out what it costs them. Usually the agent added a step, changed a part of the job they valued or got something wrong early. Fix that first. Forcing use of a tool that adds work tends to produce workarounds, not adoption. ### Should we start with the most enthusiastic team? Start with the work where a chore is clearest and the people doing it are willing to test. Enthusiasm helps, but a visible win on a job everyone dislikes convinces the sceptics more than a keen team does. ### How long does adoption take? It depends on the work and the team. A job that removes a chore and shows up in the inbox can be in use within weeks. Track human effort from the first week, so you can see whether it is going up or down. ## Sources - Fred D. Davis, ["Perceived Usefulness, Perceived Ease of Use, and User Acceptance of Information Technology"](https://doi.org/10.2307/249008) , MIS Quarterly 13(3), September 1989, pages 319 to 340. - Viswanath Venkatesh, Michael G. Morris, Gordon B. Davis and Fred D. Davis, ["User Acceptance of Information Technology: Toward a Unified View"](https://doi.org/10.2307/30036540) , MIS Quarterly 27(3), September 2003, pages 425 to 478.