Better workflows inside a restricted IT environment

I can understand why sales leaders get frustrated when an exciting AI idea meets a wall of IT approvals. The temptation is to find a shortcut. A personal account, a browser extension or a spreadsheet copied somewhere convenient. That's also how customer information ends up outside the controls your business agreed to.

Security and data restrictions are part of the job. They may come from regulation, customer commitments or perfectly sensible internal policies. I'd rather design something that can be approved and used for years than build a clever workaround that gets switched off next month.

Restrictions are part of the brief, not a reason to skip it.

Before looking for another product, find out what your business already pays for. Your CRM, document systems and productivity suite may have useful capabilities that nobody has configured or taught the team to use.

There's a much better conversation to have with IT: here's the task, here's the approved information it needs, here's where the output goes and here's how we'll check it. Can we do this safely with the tools we already have?

1. 01List what is approved. CRM, email, calendar, document storage, collaboration tools and any AI assistant already licensed. Next to each, note what it can already do that your team does not use.
2. 02Pick one task. Choose something repetitive that does not need new data access, such as preparing for meetings from existing CRM records or writing up call notes in a standard format.
3. 03Map the information it needs. Where does each piece live today? Is it accurate? Is the team permitted to use it this way? Write this down plainly.
4. 04Design the workflow inside approved tools only. Agree who checks the output before it reaches a buyer or becomes part of a record, and what good output looks like.
5. 05Bring IT in early with a one-page summary: the task, the data involved, the tools used, where outputs are stored and the human check. Ask what they would need to see to be comfortable.
6. 06Run it with a small group, against a baseline you took beforehand, for a fixed period agreed in advance.
7. 07Review time, adoption and quality honestly, then share the results with IT before asking for anything new.

Pick one task and map it properly. Who does it? How often? Which systems are involved? What information is sensitive? Who is permitted to see it? Don't skip the awkward questions because the demonstration looked impressive.

Bring IT and security into the design early. A one-page description of the workflow is usually more useful than a long list of AI features. Be clear about data flows, storage, access, retention and who owns the result.

Then run a bounded test with a baseline. Check the time saved, the accuracy of the output, the human corrections and whether people actually use it. Share the evidence before asking for permission to scale.

You don't need to become a security architect to lead this work. You do need to understand the process well enough to ask the right questions and recognise when a specialist needs to be involved.