Most mid-sized employers are being pitched AI tools with little guidance on where to point them. The usual answers are cost savings or speed. Those matter, but for many teams the real constraint is simpler: you cannot hire enough qualified people, and the ones you have are tied up in work that does not need them.
That changes the question. Automation becomes a workforce decision first and a technology decision second.
This applies well beyond insurance. It holds in finance departments, operations teams, sales support and plant offices across Canada and the United States.
Start with the people you cannot replace
Every team has a few roles where the knowledge lives in one head. A senior estimator. A controller who knows why the books are built the way they are. A scheduler who understands every customer's quirks. When one of those people leaves, the replacement takes months to find and longer to bring up to speed.
Now look at how those people spend their week. In most organizations, a large share goes to:
- Pulling numbers together from several systems
- Re-keying data from one place to another
- Chasing documents and status updates
- Preparing first drafts of routine reports
- Screening items for obvious problems before a decision is made
None of that takes years of judgement. All of it takes time from the person whose judgement you are short of. This is the best place to automate, because the gain is not only hours saved. It is senior capacity returned to the work only they can do.
Work that should stay with a person
Some tasks carry enough weight that a named human must own the outcome. Speed is not the goal there.
- Decisions with real financial, legal or regulatory exposure
- Exceptions that fall outside the normal rules
- Sensitive conversations with customers, employees or suppliers
- Sign-offs, filings and attestations where someone is personally accountable
If a regulator, auditor or major customer would expect a person to be involved, keep one involved. An AI tool can prepare the file. It should not carry the signature.
The middle ground is where most of the value sits
Most work is neither fully routine nor fully judgement. This is where AI assists without replacing anyone.
- A finance analyst gets a variance explanation drafted, then checks it and adds context.
- A sales manager gets a summary of account activity, then decides where to spend time.
- A buyer gets a flagged list of unusual purchase orders, then decides which ones matter.
- A quality lead gets patterns surfaced from inspection data, then decides on the corrective action.
The person stays responsible for the call. The tool shortens the road to it.
Four questions to sort your roadmap
Before approving any automation project, ask:
- Can we hire or train for this role fast enough? If not, it is a strong candidate for automation.
- Does the work depend on institutional knowledge or nuanced judgement? If yes, keep a person in the loop.
- What does an error cost? Expensive mistakes should stay human-led.
- Would a customer or regulator expect a person here? If so, do not remove them.
Run your most strained roles through these questions. The pattern that comes out usually looks different from a list built around cost alone.
Do not assume your team will resist
Some leaders hold back because they expect staff to push back on new tools. In our experience, resistance is usually about unclear intent. People worry the tool is a step toward cutting their job. If you say plainly that the aim is to take low-value work off their desks and that you are investing in their skills, adoption tends to go better.
That means budgeting for training, not just licences. It also means being honest about which roles will change, and moving people into the growing parts of the business where you can.
What this means for hiring
Automation does not end your hiring. It shifts it. Expect to need fewer people for data gathering and routine processing, and more who can:
- Check AI output and catch what it gets wrong
- Work with clean data and understand how it flows between systems
- Make decisions and explain them to customers and regulators
- Redesign processes around new tools
These are often hybrid profiles, an operations lead who is comfortable with data, or a finance professional who can evaluate model output. They are scarce, so plan for them early rather than after the tool is live.
What to do next
Pick your three hardest roles to fill. For each, list what share of the week goes to work that does not need the person's expertise. Automate that first, protect the judgement work, and write down the skills you will need to hire for as the work changes.
If you want a second view on which roles to build, which to automate around, and who to look for, a Traverse recruiter can help with that.