top of page

Don't Write Your AI Policy Blind

Writer: Flowstone Group
Flowstone Group
10 minutes ago
3 min read

A client asks for an AI policy template. Most people send one. The better move is to ask why.


That idea came up again and again in the security breakout at Bridgepointe's Tech Summit in Miami this month.


The message from the panel was simple: you can't protect AI you don't know about. A template won't tell you what your team is already using. We want to show why that matters, and where to start instead.



What shadow AI looks like


Shadow AI is any AI tool your team uses without approval. It is rarely a bad actor. It is usually someone trying to work faster.


  • A free chatbot used to summarize customer emails

  • An AI feature that switched on inside software you already pay for

  • An assistant a team set up on its own to draft replies or reports


Lists like these are easy to miss and seldom show up ready to read yet alone available.


What the research shows

Shadow AI is already costing companies real money. IBM's 2025 Cost of a Data Breach Report found:


Don’t write your AI policy blind. Find shadow AI first, then build rules on what’s real. Takeaways from BPT Tech Summit.

Why a template can't find it for you


A template is generic. It was written for any company, so it knows nothing about yours. It can say "only use approved tools." It can't tell you who already isn't.

Written blind, a policy can also backfire:


  • It misses the real risk. The rules cover tools nobody uses, while the risky one goes unnoticed.

  • It pushes AI further out of sight. Ban a tool people rely on with no alternative, and they keep using it quietly.


Finding shadow AI takes a different step: looking. Review which apps and AI features are in use, ask teams how they actually work, and check what data those tools can reach.


Where to start


The order matters. Start with what's real, then write the rules.


  1. Find the AI already in use. Approved or not, across every team.

  2. Trace one workflow. Pick one process. What data can the tool see, and what can it do on its own?

  3. Give every AI action an owner. A person approves what it sends or spends.

  4. Then write the policy. Rules built on what's actually running, not a template.

You don't need a full AI strategy to begin. One workflow is enough to see where the gaps are.


Don't write the policy blind


At Flowstone Group, we ask before we quote. We're vendor-agnostic, so our job is to help you see what's already running, find the risk, and match you with the right fit. Then we negotiate with suppliers so your team doesn't have to.

Could your team answer these today? Which AI tools are in use? What data can they see? Who signs off on what they do?

If not, let's walk through it together. Book a 30-minute conversation or email us at contact@flowstonegroup.com.




How we help


  • See where you’re overspending. A clear review of what you pay for across cloud, software and services, and what you actually use.

  • Get technology that fits. Solutions matched to your business, not a sales quota.

  • Negotiate with suppliers. We handle the conversations so your team doesn’t have to.

  • Bring the people to make it happen. The resources and manpower to see it through.



Sources


Comments


bottom of page