Stop running your JVs on spreadsheets, email and post-it notes. Run your JVs on Beryl.
2026 WinnerWatch Beryl read a clause.
Beryl reads and remembers every agreement, connects it to your data, and recommends your next action — the variance to challenge, the notice you owe, the cash call you're about to miss — before anyone thinks to ask.
- Pick a clause above the stage
- Hover a highlighted phrase to see what it means for you
- Press ⌘K or Ctrl K to ask it a question
Beryl reads the agreement.The words become things.Checked against the systems that hold the truth.
Illustration: Beryl reads Clause 12.3 — “12.3 — The Operator shall, within thirty (30) days of the end of each Quarter, deliver to each Non-Operator a statement of the Joint Account, and any variance exceeding five per cent (5%) of the approved AFE shall be separately identified in such statement.” — and resolves it to Ardmore Energy (Operator · 45%), Q3 2026 (Closes 30 September), AFE-1847-14 (£480,000 · 5% = £24,000), Non-Operators (Three parties · your share 18.5%). It then watches them and raises: CC-0912 is £32k over the AFE.
Beryl is a commercial control tower: continuous monitoring of every joint-venture obligation against your live financial and production data.
Illustrative — not live client data
Today's commercial reality.
Your JOA isn't a document you file. It's the control system for the asset — you've just never been able to run it.
Extract. Monitor. Respond.
Every obligation, structured.
Read once, held forever.
Every clause, deadline and threshold pulled out of your JOAs, CTIAs and TPAs, with clause references attached. A 300-page binder becomes structured data you can act on.
Checked against what's actually happening.
Continuously, not once a year.
Every obligation, cash call and AFE reconciled against your live financial and production data — the variance surfaces the day it appears, not at year-end close.
The response, drafted and ready.
You decide. Beryl does the drafting.
Notices, breach challenges and sole-risk elections drafted and cited to the clause that governs them — reviewed by a person before anything is sent.
From clause to challenge.
Take a hypothetical operator, Ardmore Energy. Here is one clause, followed from agreement to action.
Beryl reads the agreement.
Not a summary — a structured read. Every party, deadline, threshold and obligation lifted out with its clause reference still attached.
12.3 — The Operator shall, within thirty (30) days of the end of each Quarter, deliver to each Non-Operator a statement of the Joint Account, and any variance exceeding five per cent (5%) of the approved AFE shall be separately identified in such statement.
The words become things.
“The Operator” becomes a company holding 45%. “The approved AFE” becomes a number in SAP. This mapping — contract language bound to the real entities in your business — is the part that makes everything after it possible.
Checked against the systems that hold the truth.
Budget and committed spend from SAP. Production from EnergySys. The cash call itself, sitting in someone's inbox. The deadline in the calendar. Beryl joins them and does the arithmetic the clause implies.
Ask in plain English — including from Claude.
Beryl exposes your commercial estate to the assistant your team already uses. Ask Claude the question; Beryl supplies the answer, with the clause and the ledger entry behind it.
No. CC-0912 is £512,340 against an approved AFE of £480,000 — a variance of £32,340, or 6.7%.
Clause 12.3 requires any variance above 5% to be separately identified. It wasn't. Your 18.5% share is £5,983.
↳ JOA cl. 12.3 · SAP AFE-1847-14 · CC-0912Beryl drafts. Your team decides.
The challenge is written, cited and ready to send. Nothing leaves the building without a person approving it. Either way, the decision is captured as a trace in Beryl's context graph: who decided, on what evidence, and why.
Every decision becomes context.
When your team challenges a cash call, approves an AFE or declines a sole-risk election, Beryl keeps the decision as a trace in its context graph: who decided, the clause it rests on, and the evidence behind it.
- Captured as it happens, from the approval itself, not from anyone's memory.
- Connected to the clause, the cash call, the ledger and the people involved.
- Reused as precedent: the next variance starts from what your team did last time.
- Decided byJV lead, approved by Finance
- BecauseClause 12.3: any variance above 5% must be separately identified. It wasn't.
- EvidenceSAP · AFE-1847-14: £512,340 committed against £480,000 approved (+6.7%)Inbox · cash call CC-0912, received 3 Sep
- OutcomeThe operator re-issued the Q3 statement with the variance identified · 26 Sep
Why you can't just put your agreements in Claude.
Ask Claude what clause 12.3 requires and it will tell you, accurately, and you should be doing that. Ask whether you've been overcharged and it cannot know: the answer isn't in the document.
You'll still get an answer. On a £2M invoice, a confident figure tracing back to no clause and no cost line is worse than none. Beryl connects the clauses to reality:
Resolution
The agreement's terms have to be bound to real entities, numbers and dates inside your systems. That mapping is the ontology — the unique layer that Beryl brings.
Cross-agreement
A JOA clause interacts with a CTIA, and with a side letter signed three years later. One document in a context window doesn't know the others exist.
Continuity
This isn't a question you ask once. It's a check that runs every day, against data that changes every day, and raises its hand without being asked.
Precedent
Claude doesn't remember what your team decided last quarter, or why. Beryl keeps every decision as a trace in its context graph — the clause, the evidence, who signed it off — so the next call starts from precedent.
Your data never leaves your environment.
Beryl deploys inside your own cloud environment, alongside the systems it reads — no copies, no uploads, no training on your data.
Niall Rowantree
Eleven years at TotalEnergies across commercial, joint ventures and digital — running North Sea JV interests, then building the UK data and analytics function from scratch. He has been the one answering for a missed obligation, and the one asked to build the system that should have caught it.
LinkedIn ↗Pedro Santos
A decade in data and AI across the energy sector and the Bank of England. He has led data science teams inside a North Sea operator and built the production knowledge graphs and ontologies that Beryl now runs on.
LinkedIn ↗We're taking on a handful of design partners.
Not customers — partners. You bring the assets and the agreements; we build the tool around your reality. First movers set the roadmap.