Back to notes Field notes

If you work in a non-tech company, this is how AI will change in the next 2 years

Field notes •

AI strategy

Listen to summary

Download audio

If you work in a non-tech company, this is what AI will change in the next 2 years

I have spent the last five years helping non-tech companies ship practical AI. Here is how I see the next two years. AI moves into the tools your teams already use, a simple platform appears to run it safely, and governance becomes part of every buying decision.

1) Work moves faster inside the tools you already have

AI will sit inside Microsoft 365, Google Workspace, Salesforce, SAP, ServiceNow, and your industry systems. It will draft, summarize, file, check, and hand off routine steps.

What this looks like from a manager’s seat

  • Less copy-paste and fewer handoffs, so cycle times go down.
  • First drafts appear instantly, people edit instead of starting from zero.
  • Tickets, emails, and forms route to the right place with fewer mistakes.
  • Weekly numbers on time saved and error rates become the proof that matters.

What I would avoid

  • A new company chatbot that tries to do everything.
  • Ten pilots at once. Value shows up when one important workflow improves and sticks.

2) An AI platform next to your data platform

When data matured, firms adopted Snowflake, Databricks, and the cloud. AI is following the same path. Most non-tech companies will run and manage existing large models.

Your AI platform will do a few simple things well

  • Keep a catalog of approved AI tools and what they are allowed to do.
  • Store versions of prompts and rules so changes are tracked.
  • Connect to the right documents with clear citations.
  • Check quality with small test sets and show trends over time.
  • Track spend and speed per task with clear limits.
  • Keep detailed logs of who did what and when.
  • Separate test and production so changes do not break live work.

Good default stance

  • Use the smallest model that meets the target quality, use bigger models only when needed.
  • Build this on top of your existing data stack so identity and permissions stay consistent.

3) Governance will decide who scales

Boards, customers, and regulators will ask for proof that AI is controlled. In the EU this becomes concrete by 2026 and 2027, and many global firms will adopt the same standards.

What good control looks like

  • A simple inventory of AI uses, data sources, owners, and risks.
  • Clear approvals before sensitive actions or changes.
  • Evidence that quality is checked and problems are caught early.
  • Vendor terms that explain data use and retention.
  • A plan to roll back or retire an AI workflow if it underperforms.

Why it matters

  • You pass procurement faster, avoid tool sprawl, and keep pilots alive after month three.

How to tell you are on track

  • AI shows up inside tools your teams already open every day.
  • One important workflow is measurably faster and more reliable.
  • There is a single place to see quality, cost, and logs.
  • You can explain to a customer or auditor what runs, who owns it, and how it is controlled.

This is the path I use with clients. It is focused, measurable, and safe. It also matches where AI is actually going for non-tech companies over the next two years.