AI that knows
your business.

RunAgents connects AI to everything your business runs on: ERP, email, WhatsApp, spreadsheets, documents. It learns how the business actually works, then gets work done.

Talk to usSee it work
Working06:00:001,284 done today
ERPEmailWhatsAppExcelDrivePDFs
Why this exists

Most of what a company knows sits where no outsider can read it. Delivery terms live in a database field called REMARKS_2. Prices get agreed on WhatsApp. “Revenue” means one thing to sales and another to finance. Half the story of any order is in somebody's inbox. A general-purpose AI knows none of this, and it will answer confidently anyway.

So we do the dull part first: connect everything, work out what it means, and check that with the people who own the numbers. Only then does the AI go in front of anyone.

What it does

One system that knows the business, used many ways.

Illustrative data
Connected, read-only
ERP (Oracle, customised)412 tables and viewssynced 06:00
Email38 mailboxes, each with its owner's consentsynced 06:02
Shared drive6,200 documentssynced 06:05
Price lists and cost sheets140 Excel filessynced 06:05
What it has learned about this business
“Revenue” means shipped and invoiced. Booked orders are reported separately.Agreed with Finance
“On time” is measured against the first promised date, not the revised one.Agreed with Planning
Three customer codes belong to one buyer and are counted together.Found in the data, confirmed by Sales
Delivery terms are kept in a field called REMARKS_2.Found in the data
61 definitions checked with your team. 4 still waiting for an answer.
Which repeat customers ordered less this quarter, and what is at stake?
Queried sales and shipment data · 214 rows · 0.9s

Three repeat customers are down more than 30% against the same quarter last year. Over the past twelve months they placed 8.4M in orders between them.

CustomerSame quarter last yearThis quarterChange
Customer A1.9M0.9M−53%
Customer B1.2M0.7M−42%
Customer C0.8M0.5M−38%
Shipped orders, as of 06:00 today. Customer C also has an unresolved quality complaint in email from 12 days ago.
Send this list to the three account managers?
Not nowSend
Try a question:
Sales overviewRegion: North
Shipped this quarter12.6M
Orders at risk of running late5
Receivables past due1.9M
Shipped, by region
North
South
East
West
Export
Monthly shipments, last 12 months
Why is North falling?
North is down 18% on last year. Nearly all of it comes from two customers. Want the list?
Describe the job in a sentence
Every Monday, tell me which repeat customers are ordering less than last year.
triggerEvery Monday08:00
querySales and shipmentsThis quarter against last year
conditionDown more than 30%Repeat customers only
summariseA short plain noteWith the reason, where email has one
approveYou see it firstNothing goes out unapproved
deliverEmail and PDFTo you and the account managers
LiveNext run Monday 08:00Test run found 3 customers

It learns where the data lives and what your terms mean, and checks both with your team.

Read-only by defaultEach person sees only what they are allowed to seeNothing is sent without approval

Want to see this running on your own data?

Talk to us
Under the hood

Models tuned on your own data.

Large general models are powerful, but slow and costly to run on everything. Once the system knows your business, we fine-tune smaller models on your own data for the work that repeats: reading documents, sorting requests, drafting routine replies.

Routine work gets faster and cheaper per task. The large models are kept for the questions that need them.

Incoming workTuned on your dataroutine work · fast · low costLarge general modelshard questions

Ask us which of your routine work is worth a tuned model.

Talk to us
How it comes together

We build it, test it and keep it running.

Systems of recordConversationsDocuments and filesOutside dataYour business,understoodcustomerspricingproductswho approves whatAssistantsDashboardsWorkflowsfeedback and tuned models

Every rollout starts with one problem worth solving. Tell us yours.

Talk to us
Questions

What people ask first.

Why not just use ChatGPT or Claude?

They are excellent, and we use leading models underneath. What they don't arrive with is knowledge of your data, your definitions, your access rules, or anyone accountable for a workflow still running next quarter. That part is our job.

Our ERP vendor offers AI. Why not that?

If a modern cloud suite covers the whole problem, use it. We are for older, customised and mixed environments, and for work that crosses the ERP, email, files and outside data.

What do you need from our IT team?

Read-only access and one point of contact. We handle implementation and upkeep.

Where does our data go?

We agree that with you before anything is connected, including where it is hosted and which model providers are involved.

What does it cost?

The first engagement is fixed-price and built around one problem worth solving. After that it is an annual subscription sized to how much of the business is covered. We quote both before you commit to anything.

Have a question that is not here?

Talk to us
Contact

Talk to us.

Tell us where the business is losing time or money. A person will reply, and will say so plainly if we are not the right fit.

Or write to ahad@runagents.ai.

Talk to us.

Tell us where the business is losing time or money. A person will reply, and will say so plainly if we are not the right fit.