Skip to content

Guides

How to find growth opportunities with AI (November 2026)

For owners and small teams who have data in many places and no time to read it all. After reading, you can run a weekly opportunity review where every item points to its source.

Written by Growth OX4 min readUpdated: November 8, 2026

In short

  • An opportunity is a specific action tied to a specific record, not a general idea such as "post more".
  • Give the AI a small, well-labelled set of facts about the business first. Ranking improves when it knows your goals.
  • Keep only opportunities that name their source, so you can check them in seconds.

What it is

A growth opportunity is a gap between what your records show and what you want to happen: a customer who asked a question and got no answer, a product page nobody updated after a price change, a past buyer who has not heard from you. Finding them with AI means asking a model to read your records and propose such gaps, while you decide which ones are real.

How to do it

  1. Write down the goal in one sentence. "Win back lapsed customers this month" gives the AI a filter. Without a goal it ranks everything equally, and you get a long list of weak ideas.
  2. Collect the records in one place. Mail threads, order lists, support notes and past campaigns are the usual inputs. The AI cannot find a gap in a record it never saw, and splitting records across tools hides the gaps between them.
  3. Group the records by project. Client, product line or campaign works. Grouping stops the model from mixing a wholesale customer's history with a retail one.
  4. Ask for opportunities in a fixed shape. Each item needs three parts: what to do, which record it comes from, and what you expect to change. An item without a record is a guess and should be dropped.
  5. Review once a week, not every day. A fixed slot keeps the list from becoming noise. Accept, edit or discard each item and write down why; the reasons tell you what to put in the goal next week.
  6. Do the work, then measure it. An opportunity is only useful if you later learn whether it worked. See how to measure whether a growth task worked.

How Growth OX helps

Growth OX is a Mac app that follows this method. You give it a folder and it groups the contents into project areas. Each project keeps its goals, records and sources in the brain, and the assistant reads them for every answer. What it proposes appears as an opportunity card in the Today view, and the home screen and the assistant show where an item came from, so you can open the source.

Limits: Growth OX includes no AI usage. You connect your own API key, a local model or a command-line tool you signed in to yourself (see connect a model). The quality of the list depends on the records you give it and the goal you state. It proposes; it does not decide which opportunities are worth your time.

Other ways to do it

Other ways to do it
OptionGood forLimit
Manual weekly review of mail and ordersSmall volumes, full controlDepends on memory and time; gaps hide between tools
Spreadsheet with filters (last contact, last order)Clear, rule-based segmentsFinds only the gaps you thought to write a rule for
Rules in a CRM or mail toolRepeating, predictable casesRules need upkeep; new kinds of gap are missed
A general chat assistant with pasted recordsOne-off questionsYou re-paste context each time; sources are easy to lose
A project-based assistant such as Growth OXRecurring review across many recordsNeeds setup and a stated goal; you still review every item

Where to go next

Start with one project and one goal, run the review for three weeks, and keep the accepted-versus-discarded notes. Then widen the inputs. If you want to see the sorting step in detail, read how to triage customer signals.

FAQ

How many opportunities should a weekly list contain?

Few enough that you act on all of them. If you consistently discard half, tighten the goal rather than reading faster.

Can AI find opportunities without access to my customer records?

It can suggest general ideas from public information, but those are not opportunities in the sense used here. A useful item points to something in your own records.

Do I have to send my records to a cloud service?

No. A local model processes the work on your computer. If you use a cloud model, send only what the task needs and check what leaves your machine first.