The purchase order, the signed ticket and the invoice, in one file
Two sides of the same operation: the operator that contracts and the contractor that executes. The AI reads the contract, watches the order balance, validates the ticket, codes the invoice and compares each well with its own baseline. Authorizing the permit, signing the ticket and approving the spend is done by a person.
2 bases · 3 clients
Field · today’s jobs
- J-3120Latitud EnergíaRC-08
Workover · pump change
On locationticket open with no signal
OrderOC-7781 - J-3124Operadora del EsteSA-12
Pumping service
Order with balanceauthorizes mobilizing
OrderOC-7799 - J-3118Latitud EnergíaSA-19
Cleanout and stimulation
Ticket signedsupervisor and client
OrderOC-7781 - J-3130Gas del LitoralVL-02
Work at height · tower
Assignment blockedWork-at-height certificate expired on 06/10 in the proposed crew
The AI assembles the job file. Signing the ticket and approving the invoice is done by a person. · demonstration data
0
sides of the operation
operator and contractor, each with its own system
0
portals by role
leadership · field · crew · procurement · HSE
0
ticket per job
hours, consumables, photos and both signatures in one place
N0
anything that commits money
the AI proposes; authorizing and signing is done by a person
THE RULE THAT GOVERNS THE SYSTEM
The AI assembles the file behind every job and every cost. Authorizing, signing and approving is done by a person.
Prepared by the AI
Reading the contract, watching the order balance, reviewing the ticket on closing, coding the invoice, drafting the risk analysis and comparing each well with its baseline.
Decided by a person
Authorizing the work permit, signing the ticket, issuing the invoice, approving the journal entry and shutting in or working over a well.
Written down either way
Who authorized, who signed, what the AI read and which document backs each line.
THE PROBLEM
Five frictions familiar to anyone operating or serving in the field
None of them is a reservoir problem. They are all file problems: the data exists, but it arrives late and to the wrong person.
The field ticket travels on paper, or as photos in a chat
The crew writes down hours, consumables and remarks on a form. There is no signal on location, so the form comes back with the supervisor the next day, or arrives as a photo in a chat. Someone retypes it. Between what was executed and what gets invoiced there are days and, sometimes, an interpretation.
Work runs against a purchase order whose balance nobody knows
The client authorized an amount. Jobs consume it and the balance lives in a spreadsheet updated whenever someone remembers. The day it runs out the job has already started: the invoice is left hanging and everything has to be regularized backwards, with the client already annoyed.
The expired certificate shows up on site, with the crew already mobilized
Work at height, confined spaces, medical exams, the client’s induction: each with its own date. The matrix lives in a file. By the time the client’s access control turns a technician away, transport is paid for, equipment is on site and a day is gone.
Vendor invoices are coded by hand, one by one
Dozens of invoices arrive every month: chemicals, transport, rentals, well services. Someone decides which account each one goes to, with whatever criterion they remember. A new vendor is coded as best as possible, and the error surfaces at closing, when nobody remembers what that invoice was.
The well drifting from its baseline is seen a month later
Cost per barrel rises or decline accelerates, but that becomes visible once the close is ready. By then the month is gone. The conversation stops being “what do we do” and becomes “why didn’t we see it”.
THE SOLUTION
One system per side, five portals, each role seeing only its own
Leadership sees operations and cash. Field sees its work. The crew sees its ticket, even with no signal. Procurement sees contracts and balances. HSE sees who may enter today.
Users
General management, finance leadership and the controller
Access to
Consolidated and per field or base: cash, result, cost per barrel, production and contracts.
Sees on screen
A console with operations and cash on one screen, performance per asset, the invoices the AI coded and the ones it left for review, and everything waiting for their decision.
Leadership and finance
Operations console
Cost per barrel · consolidated
21.8
Wells producing
7 / 9
EXPLORE THE SYSTEM
Four modules, clickable before we talk
No screenshots: these are four modules of the operation, with demonstration data. Clients, fields, wells and vendors are fictional; the flow is the real one.
The job, the ticket and what has to clear before invoicing
The client authorizes an amount; the jobs consume it. The crew fills the ticket on location — hours, consumables, photos and signatures — and on closing the AI checks it against the contract and the order. If a check fails, the pre-invoice is held with the reason instead of going out and coming back.
- Live balance per order, with the estimated date it runs out
- A job running without an order appears before invoicing, not after
- Standby hours compared against what the agreement set
- Issuing is decided by administration, with the file in view
↓ Interactive · demo data
From the job to the invoice
OC-7781 · Latitud Energía
J-3120 · RC-08
Workover · pump change
Operating hours
11.5
Standby hours
2.0
Consumables
- Downhole packer1
- Control fluid40 bbl
6 photos with location
- Ticket signed by supervisor and client
- Purchase order with enough balancebalance remains after this job
- Rate items and prices from contract CT-441
- Standby hours within what was agreed
The AI checks the ticket against the contract and the order. Whatever does not match is flagged. · demonstration data
Who may enter a job today, and who authorizes the permit
Certificates, medicals and inductions per person, with their expiry. Planning will not let you assign someone whose certification is not current: the block appears before the mobilization, not at the client’s access gate. The AI drafts the permit’s risk analysis from the published procedure; HSE authorizes it.
- The competency matrix decides the assignment, not the memory of it
- The missing document is named in the block itself
- The risk analysis cites the procedure and version it came from
- Without an authorized permit, the job does not move to execution
↓ Interactive · demo data
Competency matrix
North base
Wílmer Ojeda
Tower operator
- Work at heightexpires 06/10
- Medical examexpires 28/02
- Client inductionexpires 15/12
Rubén Lozada
Well technician
- Confined spacesexpires 30/04
- Medical examexpires 11/01
- Client inductionexpires 15/12
Work permit PT-224
Waiting for HSE authorizationRisk analysis drafted by the AI
From the “Work at height v4” procedure and the incidents of the last twelve months.
Fall from the tower
Harness with double lanyard and an anchor point verified before climbing
Entanglement with the wireline
Delimited area and an open radio channel between operator and helper
Residual pressure at the wellhead
Verified bleed-off and gauge at zero, recorded in the permit
Authorizes: Marielena Acosta · HSE
The AI drafts the risk analysis. Authorizing the permit is done by a person, with their name and their time. · demonstration data
Every well compared with its own history
Production, cost per barrel, margin and last workover, well by well. When one drifts, the alert arrives with what the AI compared it against — its own baseline and its peer wells — and with the scenario already prepared. Shutting in or working over is decided by asset management, and the reason is written down.
- The anomaly always arrives with its baseline, never alone
- Decline compared against peer wells in the same horizon
- Marginal wells with their shut-in scenario, ready to decide
- Cost per barrel per well, not one average for the field
↓ Interactive · demo data
Assets · well against its baseline
| Well | Production | Cost per barrel | Margin per barrel | Last workover |
|---|---|---|---|---|
| SA-07Sabana Alta | 214 / 210 | 18.4 / 18.6 | 21.1 | 14/03 |
| SA-12Sabana AltaCost off baseline | 168 / 176 | 26.9 / 19.2 | 12.4 | 02/07 |
| RC-08Río ClaroDecline above peers | 141 / 182 | 22.3 / 20.8 | 16.0 | 09/01 |
| VL-05Vega LargaMarginal well | 26 / 31 | 38.7 / 36.9 | 1.2 | 20/11 |
What the AI detected
Cost per barrel drifted from its baseline over the last three weeks.
What it compared it with
Compared with its own baseline over the previous six months and with the field’s other three wells: chemical consumption and service hours rose; production did not.
The AI detects and prepares the scenario. Shutting in or working over a well is decided by a person. · demonstration data
Invoices coded with the reason and the confidence
Each vendor invoice is read and gets its account proposed, with the pattern that supports it. A vendor appearing for the first time, or a service that could go to two accounts, is handed to the controller with what the AI read — the system does not guess to look fast.
- A confidence level per invoice, visible next to the account
- Recurring patterns learned and cited, not assumed
- An account out of use gets remapped, with the change explained
- The journal entry is approved by a person, with name and time
↓ Interactive · demo data
Finance · accounting copilot
Proposed account
6105-02
Production chemicals
Confidence
96
What it read
It read the invoice number, the vendor, the purchase order cited and the product on each line.
Which pattern it cited
This vendor went to this same account in the last nine invoices, always against the same field.
The journal entry is approved by the controller, with their name and their time.
The AI codes what it recognizes and hands over what it does not. Approving the entry is done by a person. · demonstration data
WHAT WE BUILD
Eight pieces that assemble the file behind every job and every cost
They are built in the order that hurts in your operation. Each one leaves written what the AI did and what a person decided.
Purchase orders with a live balance
What the client authorized, what was consumed and what is left, up to date. When consumption crosses the threshold, the system warns and estimates when it will run out. If a job is running without an order, it appears as an exception before invoicing, not after.
A field ticket that works with no signal
Operating and standby hours, consumables, photos with their location and the signature of the supervisor and the client. It is filled in on location without coverage and syncs on return. On closing, the AI checks that hours and rate items match the contract.
Eligibility before mobilizing
Certificates, medicals and inductions per person and per position, with their expiry. Planning will not let you assign someone whose certification is not current: the block shows on the dispatch screen, with the missing document named.
Work permits with their risk analysis
The AI drafts the job’s risk analysis from the published procedure and the incident history. An HSE person authorizes it, with their name and their time. Without an authorized permit, the job does not move to execution.
An accounting copilot for vendor invoices
Each invoice arrives, is read and gets an account proposed with a confidence level. If the vendor appears for the first time or the account is unclear, it goes to review with the reason written: the AI does not guess. Recurring patterns are learned and cited.
Each well against its own baseline
Production, cost per barrel, margin and last workover well by well. When a well drifts from its baseline or declines more than its peers, the alert arrives with the comparison in view. Shutting in or working over is decided by a person.
Master agreements with watched renewals
Each contract with its expiry, its notice period and the clauses worth reviewing, extracted from the document. Automatic renewal is flagged before the notice period runs out. Where a single vendor is the only option, it is marked.
Pre-invoicing with checks before issuing
Signed tickets are grouped into a pre-invoice and run through their checks: signed ticket, purchase order with balance, contract rate. If one fails, the pre-invoice is held with the reason. Issuing is decided by a person.
WHAT IT CONNECTS TO
It integrates with whatever is already in production
Nobody changes their accounting to try out a system. What works stays and gets connected; what is missing gets built.
Claude · Anthropic
Official partnerReads contracts and invoices, drafts the risk analysis and explains every anomaly with its source
Your ERP or accounting
Integrated by API: the journal entry lands where you already keep the books, your accounting is not replaced
The client’s vendor portal
Purchase orders, invoices and payment status consulted wherever the client exposes them
WhatsApp Business API
Notices to the supervisor and the client through the channel they already use, with the file behind them
Make / n8n
Bridges to the measurement, maintenance or payroll systems already running
Supabase
Database with permissions per field, per base and per role, and a trace of every movement
TWO SIDES, TWO SYSTEMS
Walk through the one that is yours
Two systems, each designed for its own side: the operator’s management console and the contractor’s field ERP. The same rule in both, the AI prepares and the person authorizes, and the same purchase order seen from each side. Simulated AI, demonstration data.
LIVE SYSTEM · Latitud Energía
This is what it looks like inside an AI-operated system — and how it is governed.
This is not an animation: it is a demo company’s system with its modules, agents and rules. Follow the guide or browse freely.
System modules
- 01Management console
- 02Accounting copilot
- 03Assets and wells
- 04Field
- 05Contracts
- 06Equipment
Boosty Standard for Operating with AISimulated AI · demo data
LIVE SYSTEM · Latitud Servicios Petroleros
This is what it looks like inside an AI-operated system — and how it is governed.
This is not an animation: it is a demo company’s system with its modules, agents and rules. Follow the guide or browse freely.
System modules
- 01Executive panel
- 02Purchase orders
- 03Dispatch
- 04Tickets and crew mode
- 05People and equipment
- 06Pre-invoicing
Boosty Standard for Operating with AISimulated AI · demo data
THE SAME FACT · TWO COMPANIES
One purchase order, seen by the side that issues it and the side that mobilizes against it.
The balance is the same number for both. The operator sees the master agreement and how much it authorized; the contractor sees how much is left and when it runs out at today’s pace. Either system crosses over to the other.
OC-7781Latitud Energía · issues it
OC-7781Latitud Servicios Petroleros · mobilizes against it
Real screenshot of the system · demonstration data
B-O-O-S-T-Y METHODOLOGY
Four phases, one deliverable per phase, a single owner
We start where money or time is being lost today, and the rest opens up from there. The calendar is set during the diagnostic, with your team.
Discovery
We map the real flow: how a ticket arrives today, how the balance of an order is known, who authorizes a permit and how an invoice gets coded.
Deliverable
Map of the job → ticket → invoice cycle, with its break points
Architecture
Rate catalogue, competency matrix, chart of accounts, permissions per field or base and the AI policy per domain.
Deliverable
Data model, access matrix and autonomy policy
Build & Train
First the ticket and the purchase order, which is where invoicing breaks; then HSE and competencies; then accounting, contracts and assets.
Deliverable
System in production by blocks, with its audit trail
Adopt & Scale
We train crews, supervision, procurement, HSE and administration on their own portal. New bases, fields or clients are added afterwards.
Deliverable
An autonomous team that opens a new base or field without us
B-O-O-S-T-Y METHODOLOGY
Four phases, one order, one owner per task
This is how an end-to-end implementation is chained: every bar is a concrete task, with the role that owns it, and the marker runs through the sequence. The calendar is set in the assessment, with your team and your timing.
F1
Discovery
F2
Architecture
F3
Build & Train
F4
Adopt & Scale
Frequently asked questions about oil and gas systems
For both, and they are different systems. The operator needs assets, field, vendor contracts and the accounting behind all of it. The contractor needs contracts and rate items, client purchase orders, dispatch, tickets, HSE and invoicing. That is why they are two systems with their own design, each built for its side, with the same rule: the AI prepares, the person authorizes. You can open both on this page and cross from one to the other over the same purchase order.
Yes, and that is a requirement, not an extra. The ticket is filled in on the phone on location without coverage: hours, consumables, photos with their location and the signatures are stored on the device and upload by themselves when signal returns. On sync, the AI checks that what was recorded matches the contract and the purchase order, and whatever does not match is flagged for the supervisor.
No. The AI detects that a well drifted from its baseline, builds the scenario with the comparison and leaves it ready; shutting in or working over is decided by asset management. The AI drafts the permit’s risk analysis from the published procedure and the history; an HSE person authorizes it, with their name and their time. Anything touching committed money or somebody’s safety stays a proposal.
Not necessarily. If your accounting works, it stays: the entry the copilot proposes lands by API where you already keep the books. What gets built is what today lives in spreadsheets and chats: the ticket, the order balance, the competency matrix, the permit, the well file. What is integrated and what is built is decided during the diagnostic, looking at what you already have.
We work through whichever channel does exist: email, the client’s form or the document they return. The system keeps the purchase order registered with its balance, reads the document when it arrives and keeps the original as evidence. If a portal or API appears tomorrow, it connects without changing your people’s flow.
Yes. Bases or fields in different countries operate with their currency, their tax and their issuing format, and leadership sees the consolidated view. Each country’s tax requirements are defined during the diagnostic, because they change per jurisdiction and per taxpayer type; the system is built around those requirements, not the other way round.
It is set per role and per scope. The crew sees its job, not the margin. The supervisor sees their field or base. Procurement sees rates and contracts. Leadership sees the consolidated view. The client, if a portal is enabled for them, sees the progress and documents of their jobs, not your cost. Every view and every change is logged with user and time.
It is defined during the diagnostic. It depends on how many fields or bases come in, how many clients or operators have to be modelled, which systems it must live with and whether the ticket today is paper or already digital. We leave the diagnostic with the scope written down, the order of the blocks and what would be measured in each one. No invented ranges before looking at your operation.
A WORD FROM THE FOUNDER
“In the field the problem is almost never the well. It is the form that came back on Thursday and the purchase order nobody knew was exhausted.”
I have seen good crews, equipment in good shape and supervisors who know their trade, arguing with a client over a two-month-old invoice because the ticket said one thing and the contract said another. And I have seen a full mobilization turned away at access control because a technician’s certificate had expired in a file nobody opened.
None of that is a technical problem. It is a file problem: the order in a spreadsheet, the ticket on paper, the certificate in a folder and the invoice on somebody’s desk. That is exactly where AI helps: reading the contract, watching the balance, reviewing the ticket on closing, coding the invoice and comparing the well with its own baseline. Authorizing the permit, signing the ticket and approving the spend stays with a person, and that is not a technical limitation: it is the system’s policy.
If you operate fields or provide field services, book thirty minutes with me. Not with a sales rep. With me. We leave with the map of your job → ticket → invoice cycle and which block to build first.
Gabriel Montiel
CEO · Boosty Digital
GET STARTED
Ready for the ticket and the purchase order to stop living in a spreadsheet?
Book a 30-minute diagnostic. We review how a ticket arrives today, how the balance of an order is known and who authorizes what. We tell you what we would build first and how it would be measured.