Build AI your team can rely on, one column at a time.

Give your team a layout they already know, so they can build and fine-tune without waiting for an engineer.

F2fx

Is safe to approve? Check its , the date and our .

	B	C	D	E	F
1	Supplier	Certificate	Expiry	Policy	✦ Decision
2	Northwind Foods	BRCGS food safety	2027-03-31	Supplier policy v3	Approve valid, all sites covered
3	Harbour Dairy	ISO 22000	2026-11-02	Supplier policy v3	Approve valid until November
4	Fenwick Mills	Page 2 missing	unknown	Supplier policy v3	Review certificate incomplete
5	Alder & Sons	FSSC 22000	2025-12-15	Supplier policy v3	Reject certificate expired
6	Coastline Produce	BRCGS food safety	2027-06-30	Supplier policy v3	Approve valid, all sites covered

Build every step where you can see it.

DAGL is named after the four parts of everything you build in it.

DData	Your records, one per row: suppliers, invoices, tickets, contracts.
AAgent	The AI steps, one per column: fetch a file, read the facts out of it, ask the question, send the answer on.
GGrid	Every step's result on every record, side by side, so anyone on the team can see what happened.
LLoop	Compare the answers with your expert's, fix the step that went wrong, and run it again as a new version.

Start with work your team does by hand today.

Each one becomes a check your team runs every week, with the working kept beside every answer.

✦ New column

✦Ask a question about each rowCheck every supplier certificate and flag the ones that have expired.

✦Pull fields out of textPull renewal dates and notice periods out of every contract.

✦Sort rows into categoriesSort every ticket by topic and urgency, and draft a first reply.

✦Rank rows by your criteriaRank companies or applications on your own criteria, with the evidence for each kept.

✦Match one list against anotherMatch invoices to orders and query the ones that do not fit.

✦Research and score each rowResearch every lead and score how well it fits.

Know when you can rely on it.

DAGL runs each new check on 50 cases your expert has already answered, and counts the matches. Fix the instructions where they differ and run those cases again. When they agree, run the check on everything.

46 of 50 cases agree with your expert

1	Supplier	✦ Decision	Your expert	Agree
2	Northwind Foods	Approve	Approve	✓ Agrees
3	Harbour Dairy	Approve	Review	✗ Differs
4	Fenwick Mills	Review	Review	✓ Agrees
5	Alder & Sons	Reject	Reject	✓ Agrees
6	Coastline Produce	Approve	Approve	✓ Agrees

Rows 7 to 51 are below. Every row has an answer from both. example data

Use it where answers have to hold up.

See the working.

Every answer shows what it read, the instructions it followed and the version that produced it.

Know the cost first.

Each run shows how many rows it will read and what it will cost, before anything is sent.

Keep every version.

Compare versions on the same cases, and go back whenever you need to.

Keep your data with you.

Run DAGL on your own servers and use your own AI account.

Let everyone work the way they already know.

The expert

Reads real cases in rows and columns and improves the instructions in plain words.

The builder

Connects the data and sets up each step, in the grid or from code.

The owner

Sees the cost and the agreement count, sets the budget and approves what goes live.

Prove to every client that the AI does the job.

Give clients a check they can use while you keep control of it. Build it once and copy it into each client's workspace. Each client sees how often it agrees with their own team.

Talk to us about partnering

supplier_check · copied to 3 clients

Client	Cases	Match their answer
Client A	50	48 of 50
Client B	32	32 of 32
Client C	75	71 of 75

example data

Bring one check your team does by hand.

Try it on your data