The AI Operating System for the oil & gas industry.
One system schedules every inspection task, reads every report, flags what's critical, and follows up on each hand-off between Inspection, Maintenance and Operations until it is closed with evidence, with your engineers deciding, not chasing.
A refinery doesn't get hurt by what it finds. It gets hurt by what it drops.
Every finding must cross three departments before it is fixed. Today that journey runs on notifications, emails and memory, and no single system owns it.
Where the loop lives todaysix systems, none of them talking
Carried by hand, per finding
One finding's journey
Five hand-offs, four desks, nobody owning the relay.
01 · Inspect & type up — Findings hand-written on the round, keyed into SAP later.
Inspect & type up
Inspection
Read & notify
Inspection
Assess & decide
Maintenance
Approve & budget
Operations
Chase & report
Leadership
Months from finding to closure, while remaining life ticks down.
Follow one finding from flag to close-out.
The loop that takes months of chasing today, run by the system: the task creates itself, the gate fires, the follow-up is chased daily, and a stalled decision escalates before remaining life is spent.
Finding FND-2214 · CU-OH-04
CU-OH-04 · crude overhead
The inspection task creates itself
The RBI model puts CU-OH-04 in the window. It arrives with an owner, a date and acceptance criteria — nobody typed it.
Two follow-ups, one escalation, zero dropped hand-offs — and nobody built a report about it.
One governed loop, from inspection plan to closure.
Not a dashboard bolted onto SAP: a staged workflow where AI schedules, reads, chases and escalates, and your people make every decision that matters.
Turn the plan into scheduled tasks
The RBI inspection plan, meeting actions and programme milestones become dated tasks with named owners, created by the system and dispatched over WhatsApp. Nobody types a task.
- RBI windows become inspection rounds with owners and due dates
- Tasks arrive on WhatsApp as a link: no login, no new app
- Meeting minutes and annual milestones feed the same schedule
Turn the plan into scheduled tasks
The RBI inspection plan, meeting actions and programme milestones become dated tasks with named owners, created by the system and dispatched over WhatsApp. Nobody types a task.
- RBI windows become inspection rounds with owners and due dates
- Tasks arrive on WhatsApp as a link: no login, no new app
- Meeting minutes and annual milestones feed the same schedule
Read every report, extract every finding
Inspectors upload reports from the field; Document Intelligence reads them and pulls out readings, mechanisms and findings, with every value pointing back to its source page.
- UT readings, corrosion rates and damage mechanisms extracted automatically
- Every value keeps a pointer to the exact page it came from
- Scanned, photographed and messy documents normalised
Flag critical findings the moment they land
A gate checks every finding against your rules, remaining life against the next turnaround, measured wall against T-min, and fires the SAP notification itself, evidence attached.
Rule: remaining life < time to next turnaround
- Remaining life vs time-to-turnaround checked on every finding
- Notification raised in SAP with readings, photos and asset history attached
- No human in the hand-off to forget it
Chase the hand-off, escalate the stall
The assessment is chased daily and risk-scored continuously. If the decision stalls past its date, it escalates up the ladder automatically, before remaining life is spent waiting.
- Repair-now vs defer-to-turnaround chased to a decision
- Risk climbs as remaining life is spent, not when a timer expires
- Stalled hand-offs raised to leadership without anyone complaining
Route money and change to humans
Budget goes to the asset owner, process change goes to MOC in Intellect. AI routes, chases and assembles the evidence; a named person always approves.
- Budget requests routed to the right approver with the case attached
- MOC raised and tracked in Intellect, never bypassed
- AI never approves spend or change: it keeps the request moving
Execute, verify and close with evidence
The work order is scheduled into the window in SAP; on completion, NDT, cost and records are gathered onto one close-out screen and leadership signs once. The dossier archives itself.
- Work order created and scheduled in SAP by the gate passing
- NDT verification, cost and integrity record on one close-out screen
- Reporting is a by-product: the Eagle View is always current
One system, three surfaces.
Everyone works on the same engine, but each person sees only the view that answers their question — the inspector's phone, the team lead's board, and leadership's Eagle View.
Inspector
“What do I do next?”
CU-OH-04 survey due today. Open your round: 9os.link/r/118
Readings extracted ✓ Round complete. Nothing else today.
- Tasks arrive on WhatsApp — no login, no new app
- Checklist and photo upload from the field
- Readings extracted automatically, no re-typing
Team lead
“Where is my work, and who is carrying it?”
- The Task Schedule OS: every task, owner and date
- Follow-ups chased and escalations raised for you
- Workload visible, so it can be shared fairly
Leadership
“What needs my attention?”
- One live Eagle View — the ~20 hand-built reports retire
- Escalations and stalls surface themselves
- Branch health and workload at a glance
One engine underneath
Reporting stops being a job. It becomes a by-product.
Nobody fills in a dashboard. Because every task, hand-off and approval runs through the system, the picture assembles itself, and three chronic failure modes disappear.
~20 hand-built reports, always slightly out of date
One always-live Eagle View for leadership
Hand-offs between departments get dropped
Every hand-off tracked end to end, chased daily
Nobody can see who is overloaded or what has stalled
Workload and risk visible per person, escalated automatically
The same 9OS engine already runs a SAP-integrated plant
Read the case studyA SAP-integrated manufacturer runs its whole quoting workflow on this core: turnaround 92% faster, ₹15L/month saved, 585% first-year ROI. The oil & gas configuration runs the same engine — workflows, gates and agents — over your inspection, maintenance and operations loop.
How it works under the hood.
Four agents, each with one job, over a gate engine. And because Workflow → Stage → Gate → Task is configuration, not code, a new loop is a config change, not a rebuild.
Document Intelligence reads the paperwork
Inspection reports, corrosion surveys, drawings and meeting minutes parsed as they land. Findings, readings and actions are extracted with a pointer to their source, nothing re-typed, nothing invented.
Scheduler turns plans into tasks
The RBI plan, gate outcomes and meeting actions decompose into dated tasks with owners and dependencies. Completing one stage creates the tasks for the next, so the schedule maintains itself.
Project Manager & Escalation keep it moving
Daily priorities set, open hand-offs chased, every task risk-scored continuously. Exceptions route up the ladder automatically, even when nobody wants to speak up.
SAP, Meridian and Intellect stay in charge
Notifications and work orders are raised and read in SAP, MOC stays in Intellect, documents in Meridian. 9OS runs the loop across them; your systems of record are untouched.
Built for a plant, not just an office.
An operating layer inside a refinery is held to a different standard. The system is self-hosted, fenced off from plant controls, and every decision that matters stays human.
Never touches plant controls
A visibility and coordination layer only. It never connects to DCS, SIS or any safety-instrumented system: it coordinates people, tasks and records, not the plant.
Self-hosted, in-country
Deploys inside your own environment or private cloud, built to refinery cyber-security and data-residency requirements. Inspection data, costs and records never leave your servers.
Humans hold the pen
Money and safety decisions always stop at a named person; AI never approves spend or change. Every AI action is logged and explainable, nothing is a black box.
We deploy it on your floor and own the outcome.
We don't drop software and leave. A forward-deployed engineer learns how your business actually runs, deploys the system on your real work, and drives it into the team, shipping working AI every month.
Onboarding Audit
Context & PlanningA forward-deployed engineer maps people, process and data across your branches, finds the one loop worth automating first — usually inspection-finding-to-closure — and sizes the value. You keep the playbook either way.
Explore the Enterprise AI AuditEmbedded AI Implementation
Implementation & AdoptionWe run one real loop live for one nominated team: SAP connected, gates and the escalation ladder configured, tasks arriving on WhatsApp, and a real Eagle View replacing that team's reports.
Explore Embedded AI ImplementationAI Operating System
Run & OptimizeOnce it's live and adopted, it becomes your operating system on an annual license, with an engineer who keeps it sharp, retrains it on your data, and keeps shipping — branch by branch, then site by site.
Explore the AI Operating SystemThe Enterprise Audit is free, and you keep the playbook either way.
The questions every technical services head asks first.
Never. Gates route money and safety decisions to named people: the asset owner approves spend, the MOC board approves change. The AI assembles the evidence, chases the thread and escalates the stall; a human always signs, and every AI action is logged and explainable.
Take every finding from flag to
verified closure.
Bring us one live finding. We'll walk it through the loop on a call, and show you your own plant running on the system.
The first step is always a conversation.