Decision trace infrastructure
Stop starting every decision from scratch.
approve
Septon helps your teams make decisions from your own data and conversations, and records each one with the evidence behind it, at the moment it is made. The next decision starts from what your company already knows, and every decision can be explained, improved and defended.
Claude
ChatGPT
Copilot
Slack
TeamsDecisions made in a company, right now
Examples. Almost none of them leave evidence behind.
The way in
30 days to your first decision on the record.
A working result, not another slide deck.
- Week 1
Leadership interviews
We learn how decisions really get made in your company.
- Week 2
A working session with your team
We map your decisions by layer, count how many are made by people and how many by AI, and choose the one to put on the record first.
- Day 30
One real decision, live in your ledger
Running end to end with your own data, plus your value-ranked big bets and a people and training plan.
01 / The problem
Three questions about your company.
Answer them for your own business. Together they show where your decisions are made, what they are worth, and whether you could show why they were made.
Question 1 of 3 · Decision types
How many decisions is AI making in your company right now?
Every decision sits at one of four altitudes. Stakes fall and volume explodes as you go down the stack.
The volume lives exactly where the evidence doesn't.
LAYER 1
Strategic
a handful a year
Where to play. What to bet on. Rare, irreversible, board-level.
- Enter the German market
- Acquire a competitor's data business
LAYER 2
Allocation
hundreds a year
Where money, people and capacity go. Quarterly and annual rhythm.
- Move £2M from events to retention
- Add four engineers to the platform team
LAYER 3
Operational
thousands a week
How the business runs, week to week, owned by the functions.
- Reprice the top 50 accounts for Q4
- Switch packaging to a second supplier
LAYER 4
Transactional
millions a day
Made at the frontline or by systems, all day, every day, and increasingly by AI, not people.
- Approve motor claim 4418
- Discount 8% on trade order 55102
Right now, somewhere in your company, a claim is being approved, a discount granted, a customer declined. Ask who decided, and on what basis. Almost nobody can show you.
THE PUNCHLINE
A handful of strategic decisions a year. Millions of transactional ones a day.
Your governance lives at the top of the stack. Your volume lives at the bottom, and so does your AI. Almost nothing there leaves a trail.
Question 2 of 3 · What's at stake
How many of them are worth more than $10k?
Scrutiny follows the million-dollar decision. Multiply each layer's value by its volume and the money flows through the bottom of the stack, unexamined, at machine speed.
Run it on your own numbers: count the decisions in each layer, multiply by what one is worth, and you've priced what flows through your business unrecorded.
Question 3 of 3 · The evidence
Could you show the evidence behind those decisions?
Everything around a decision is captured: the emails, the meetings, the data. Septon records the decision itself, with the evidence behind it, at the moment it is made. Choose an industry to follow one decision, step by step.
What should our Q4 discount policy be for the top 50 accounts?
Julia, the VP Commercial, asks in Claude. Septon brings in the context around the question.
Seven sources, last quarter's decision, an 18% margin floor, eight contract holds and $4.2M a year in discount leakage. Saved as it stood that day.
Three options were weighed. Julia chose to phase the rollout, 42 accounts now and 8 in January, and wrote down why.
Finance reviews the six margin-risk accounts. Each action has an owner and is ticked off as it is done.
A year later, anyone can open this record, see why it was decided, and learn from it.
Should we raise the credit limit for a long-standing business customer?
A relationship manager asks in Copilot. Septon brings in the context around the request.
The customer's payment history, the bank's lending policy, two earlier exceptions and the current exposure limit.
Raise it in two steps, hold it, or decline. The manager chose two steps, and the reason is on the record.
Credit risk reviews the second step in 90 days. The customer is told, and the limit is updated once it is approved.
If a regulator or the customer ever asks why, the answer is one click away.
Which pharmacies come first when a medicine is in short supply?
A supply planner asks in Claude how to share out a limited delivery.
Stock levels, open orders, patient-critical flags and last month's allocation rules, as they stood that morning.
Three allocation rules were weighed. The planner chose to protect patient-critical orders first, and wrote down why.
The warehouse gets the pick list, the affected pharmacies are told, and a review is set for the next delivery.
The next shortage starts from this decision, not from a blank page.
Should we move one production line to a second packaging supplier?
A plant manager asks in Copilot. Septon brings in the context around the question.
Quality records, delivery times, price quotes and last year's supplier failures.
Stay, split or switch. The plant manager chose to split the volume for one quarter, with the reasons recorded.
Procurement renegotiates, quality sets up a test, and the plan is reviewed against real results at the end of the quarter.
When the supplier question comes back, the history is already there.
Should we ship this feature to every customer, or to a small group first?
A product lead asks in Claude whether to release a new feature to everyone or to a small group first.
Test results, support tickets from the pilot, the release checklist and the last two incidents.
Everyone now, 10% first, or wait. The lead chose a 10% release for a week and wrote down the risk that decided it.
Engineering sets the rollout flag, support is briefed, and the team reviews the numbers on day seven.
When something goes wrong later, the reasoning is already on the record.
How should we share a limited seasonal product across our stores?
A regional planner asks in Copilot how to split a limited seasonal product across the region's stores.
Last season's sell-through, current stock, weather forecasts and each store's sales so far.
An even split, follow past sales, or favor the busiest stores. The planner chose past sales with a reserve, and recorded why.
Stores get their allocations, the warehouse updates the picks, and the plan is checked against sales after two weeks.
Next season's planning starts from what worked.
Which carrier should take this week's delayed shipments?
A dispatcher asks in Teams which carrier should take a batch of delayed shipments.
Carrier rates and delivery times, the customers' deadlines and recent delivery failures.
Stay with the usual carrier, split the batch, or pay for express. The system proposed a split and the dispatcher confirmed it.
Bookings are made, customers are told the new dates, and the extra cost is logged against the delay.
Every rerouting decision can be explained to a customer.
Should we start planned maintenance now or after the cold spell?
An operations manager asks in Copilot whether to start planned maintenance now or wait for the cold spell to end.
Weather forecasts, demand projections, equipment condition reports and the contract penalties.
Start now, wait a week, or do it in two stages. The manager chose two stages and recorded the safety reasoning.
The crew schedule is updated, the grid operator is informed, and the plan is reviewed when the first stage ends.
The next outage decision starts from this one.
How should we shift the campaign budget this quarter?
An account director asks in Slack how to move budget between two campaigns.
Spend and results to date, the client's targets and the notes from the last review.
Hold, shift a quarter, or shift half. The director chose a quarter and noted the client's concern.
The media team updates the plan, the client gets a summary, and results are checked in four weeks.
Next quarter's budget conversation starts from what was decided and why.
Bring in the people who need to review it.
Send a decision to Legal, Finance or Risk before it is final. They review it inside the record, and can see how they ruled on similar cases before.
Legal's earlier decisions on similar cases
- Q2Tiered discount bands for 12 accountsApproved
- Q1Early-renewal discount on a contract holdApproved with edits
- Last Q4Price lock for a key accountDeclined
Once it is on the record, it keeps working.
Every recorded decision teaches the next one, and starts the work that follows.
When the decision is made
It starts the work in the tools you already have.
Septon hands the decision, with its context, to your own agents and workflows. Nothing waits on someone remembering.
Re-run the margin check on the six at-risk accounts.
In your workflow tool, with an owner and a due date.
A message in the chat tool they already use.
In the system of record, once approved.
Each action reports back to the record, so you can see what was decided, what was done and what happened next. This is an example flow.
02 / What we do
We don't start with software. We start with your decisions.
Septon is a partnership and a platform, built by people who ran enterprise data and AI inside McDonald's, Walmart and Mars. Six services, one platform underneath.
AI strategy and ROI modeling
Board-level work from leaders who ran enterprise data and AI at McDonald's, Walmart and Mars. Decide where AI makes you money, size the return, rank the big bets.
Decision audit
Map your most important decisions by level, shape, cost and decider. Then see how many are already AI-touched with nothing behind them.
AI capability building
Executive sessions plus practitioner training that gets teams using AI properly in their day jobs.
Custom agent development
Agents built for your real workflows, with every decision they make signed to your ledger as evidence.
Forward deployed engineering
Our engineers embed with your team to get AI working inside your operations, not in a sandbox.
AI governance and EU AI Act readiness
A structured assessment of where you stand against the obligations now being enforced, and what to fix first.
Underneath all of it sits the Decision Ledger.
It seals the record of every decision, builds your company's brain from what you decide and learn, and completes the actions each decision triggers. You keep it, and it compounds.
Start with 30 days
Week 1 leadership interviews. Week 2 working session with your team. By day 30 value-ranked big bets, a people and training plan, operating model changes, and one real decision running end to end in your own ledger.
A working result, not another slide deck.
03 / Why Septon
Judge us the way you'd judge any investment: by the outcome, and by when it arrives.
Defend
Pull the full record of any decision in minutes, not weeks of reconstruction from screenshots and logs.
weeks of reconstruction → minutesLearn
See what worked and stop repeating what didn't. The rationale stays when the decision-maker leaves.
your next decision starts from the last oneAct
The decision triggers the actions it requires, so nothing waits on someone remembering.
days of follow-through → secondsHowever your organization would usually tackle this, the comparison is the same:
04 / The compliance spine
Nine obligations. One evidence layer underneath all of them.
We structured Septon around the nine core obligations the EU AI Act places on high-risk AI, so every record you capture maps directly to what you must demonstrate.
Show the nine obligations
OB-01Risk management systemArt. 9OB-02Data and data governanceArt. 10OB-03Technical documentationArt. 11OB-04Record-keeping and loggingArt. 12OB-05Transparency to deployersArt. 13OB-06Human oversightArt. 14OB-07Accuracy, robustness, securityArt. 15OB-08Quality management systemArt. 17OB-09Conformity and registrationArt. 43Compliance is the floor, not the point. The same evidence that satisfies the record-keeping and human oversight rules is what lets commercial teams move faster, because defensible decisions are decisions you can delegate.
05 / About us
The experience behind Septon.
Septon is built by operators, data scientists and engineers who have done this work inside some of the world's largest companies. Between us we have led global data science and the Customer Data Platform at McDonald's, run engineering and data science teams at Walmart, spent thirteen years at Mars, built forecasting platforms for Tata Steel, and led AI transformation work for Heineken and Saudi PIF.
We bring that experience to one problem: making every decision explainable, improvable and defensible.
Experience at and for
06 / Questions leaders ask
Asked in every boardroom. Answered here.
Can't we just build this ourselves?
You can log; logging isn't evidence. By the time you've built tamper-evident sealing, policy mapping, oversight capture, obligation coverage and audit-grade exports, and kept them current with the EU AI Act, you've built Septon, slower. The trade is time and focus.
Is this just EU AI Act compliance?
Compliance is the floor. The same records that satisfy the nine obligations are what let you delegate more to AI, and what compound into institutional judgment. Nobody should buy a ledger twice: once for the regulator, once for the business.
We already have model monitoring.
Keep it. Monitoring watches models: drift, accuracy, bias. Septon records decisions: who decided, on what basis, with what oversight. When a decision is challenged, a dashboard of model metrics is not an answer.
Will this slow our decisions down?
The opposite. Capture happens at the point of action, not as an extra step. And decisions speed up overall, because a defensible decision is one you can delegate, downward and to AI.
What's the engagement model?
Thirty days to a working result: leadership interviews in week one, a workshop with your team in week two, and by day 30 one of your real decisions running end-to-end in your ledger. Then an ongoing partnership, if the result earns it.
Do our people have to change tools?
No. Decisions stay in the tools your people already use: Claude, Copilot, your existing workflows. Septon records them where they happen.
07 / Talk to us
Your company's decisions are its most valuable untapped asset. Start putting them to work.
Write to us
Prefer email? Send a note and we will get back to you.
A working session starts with one real decision from your business.