Slack AI assistant

Ask your business a question in Slack and get the real number back

Owners and teams ask the same questions every week: how did we sell last week, how many weeks of stock are left, which invoices are due. Someone opens three tools and a spreadsheet to answer. A Slack assistant can answer in seconds, but only if the AI never does the maths itself. The figures come from fixed calculations over your data; the AI picks the right one and explains it.

Sounds familiar?

Where the hours go today

  • Simple questions about sales, stock or cash wait for the one person who knows where the data is.
  • Numbers live in Shopify, Xero, the CRM and a few spreadsheets that don't agree with each other.
  • The Monday report takes half a day to put together by hand.
  • A general chatbot was tried and it confidently made up figures.

How it works

The same job, before and after.

Questions answered with figures you can check

A question in Slack is routed to one of three paths: a number question goes to a fixed calculation over your data, a how-to question goes to your documents, a find-me question goes to search. The answer shows the figure, how it was calculated, the source and when the data was last updated.

By hand today

  1. Question in Slack
  2. Someone opens three tools
  3. Builds the number in Excel

With automation

  1. Question in Slack
  2. AI picks the right calculation
  3. Code computes the figure
  4. Answer with source and date

A Monday briefing that writes itself

Every Monday at 8:00 the owner gets a short post: last week against the week before, anything unusual, invoices and deliveries coming up. Same calculations, no one has to assemble it.

By hand today

  1. Half a day on Monday
  2. Copying into a report

With automation

  1. Monday 8:00
  2. Briefing written from the figures
  3. Posted in Slack

What you get

  • A Slack app with your name and avatar: DMs, mentions and threaded replies
  • A data hub (PostgreSQL) fed from Shopify, Xero, Amazon, your CRM or spreadsheets
  • Metric definitions written as code, agreed with you, never improvised by the AI
  • Answers from your documents in Notion, Google Drive or SharePoint, with links
  • Role-based access, for example finance answers for founders only

Built so it doesn't embarrass you

  • The AI never calculates figures; it only chooses and explains calculations you approved.
  • Every answer says where the number came from and how fresh it is.
  • If the data is stale or a question is ambiguous, the assistant says so instead of guessing.
SlackMicrosoft TeamsShopifyXeroQuickBooksAmazon Seller CentralHubSpotNotionGoogle DrivePostgreSQLAnthropicOpenAI

Questions

Frequently asked

Why not let the AI read the data and answer directly?

Because language models are unreliable at arithmetic and will happily produce a plausible wrong number. Here the maths is done by tested code; the model decides which calculation fits the question and puts the result into words.

Does it work in Microsoft Teams?

Yes. The same assistant can run as a Teams app; the data hub and calculations stay the same.

Where is our data stored?

In a database you own, hosted in the EU or UK if you need that. Model providers can be used with zero data retention, or through AWS Bedrock in an EU region.

Not sure this fits your business?

Answer eight short questions about how your week looks. Within two working days you get a written answer: what can be automated, how it would work and roughly what it would cost. Free, no call needed.

Start the free check

Other automations