The AI you already pay for. Already have ChatGPT, Claude, Gemini or Copilot, but your team uses it to write emails, not to run operations? We connect it to your systems, your data and your rules, with the permissions and review you define.
Hábil AI
Your AI project doesn't fail because of the model. It fails when it has to connect to your ERP, your CRM, your core and your WhatsApp. We've been doing that since 2006.
We connect the AI you already pay for, or models on your own infrastructure, to what you already run, with the permissions, rules and human review your team defines. The AI proposes and leaves a trail; your system and your people decide.
Feasibility review on WhatsAppMost companies have already tried AI; few have it working in their daily operations. The model is rarely the problem: the problem is connecting it to your systems, your rules and your people without putting what already works at risk.
Use the AI you already pay for, or one that runs in-house
AI on your own infrastructure. When conversations and documents must be processed inside your company, not in a provider's cloud, the models run on your servers or in your private cloud. Before you invest in hardware, we measure which model fits and at what speed.
Both at once. What's sensitive is processed inside your company; everything else uses the AI you already pay for.
Where AI already pays off
Finance
VAT you don't lose
You pay invoices whose payment complement (complemento de pago) never arrived, and you can't credit that VAT. Nobody knows how much it is.
You know how much VAT is at risk, from which supplier and on which invoice, in time to recover it.
It was hard because you have to match what the SAT (Mexico's tax authority) says against what your banks and your accounting say, invoice by invoice.
We integrate e-invoicing and SAT data with our clients' accounting systems.
Technology
AI on top of your systems, without touching your core
When an interface between your ERP —SAP, Oracle, Microsoft Dynamics, Odoo, CONTPAQi or another—, your CRM and the rest of your operation fails, your people spend hours reading technical logs to figure out what broke.
The AI tells you in plain English what happened, which invoice, payment or purchase order it affects and what to do. Your system is still the one that decides.
It was hard because you have to understand, at the same time, the system that fails, the one that receives and the business behind them, without putting the operation at risk.
We connect the most widely used ERPs and CRMs —SAP, Oracle, Microsoft Dynamics, Odoo, Salesforce, HubSpot— with the operations of banks, insurers and regulated companies.
Operations
Zero data entry
Your team spends hours copying data from IDs, tax certificates, proofs of address, bank statements and invoices into your systems.
The data arrives on its own, with evidence of where it came from, and a person only confirms it.
It was hard because documents arrive as photos, crooked and incomplete, and a wrongly captured piece of data becomes a compliance problem.
We've already solved ID reading for a bank in Mexico.
Banking and fintech
Account opening without a branch
Opening an account meant going to a branch, and someone typed in the data from each ID by hand.
Your customer opens their account from their phone: we check that a real person is in front of the camera, compare their face with their ID and validate their data against the INE (Mexico's voter registry) and RENAPO (the national population registry). Doubtful cases go to a person on your team.
It was hard because photos arrive crooked and incomplete, and every validation has to withstand an audit.
We built the identity validation for a bank in Mexico, including the INE lookup; today we offer it complete.
WhatsApp collections that close with a payment
Your collectors call and send messages that go unanswered, and when the customer does reply, nobody has the balance or the payment options at hand.
The AI replies with the account's real data, negotiates within your rules, sends the payment link and confirms the payment; anything sensitive goes to a person.
It was hard because every reply has to come from your system, within your policies and without risking a complaint.
We built direct-debit collection for an insurer in Mexico.
Insurance
The simple claim, ready to decide
A simple claim sits on the desk for days while someone gathers documents and reviews the policy.
The case reaches the claims adjuster complete, and the decision is still theirs.
It was hard because every policy's rules are different, and a mistake costs money and reputation.
We build policy issuance and claims systems for insurers in Mexico.
Claims reporting on WhatsApp, 24 hours a day
Policyholders want to report at any hour, and your claims desk receives loose photos and documents through several channels.
The policyholder reports, sends photos and documents and checks the progress from the chat; your desk receives the file in order.
It was hard because every report has to reach your claims system complete, without manual data entry.
We built the mobile apps that policyholders of two insurers in Mexico use to handle their requests.
I head up…
| Role | Your question | Takes you to |
|---|---|---|
| General Management | Where does AI pay me back this year without putting operations at risk? | |
| Technology (CIO / CTO) | I already have an AI contract. How do I use it in my operations? | |
| Sales | How do I keep the AI from promising what it shouldn't? | |
| Operations | How do I stop manual data entry? | |
| Finance | How much money is slipping away from me? | |
| Collections | How do I collect without risking a complaint? |
Control, security and your data
What the AI never does on its own
- It doesn't collect, approve, issue or modify a record unless a person or a rule you defined authorizes it.
- It doesn't offer prices, discounts or terms that aren't in your rules.
- It only has the permissions your Technology team gives it, per system and per action.
- Every proposal is logged: what it saw, what it proposed and who authorized it.
- If it makes a mistake, it's corrected like any other change in your system, because the action is executed by your system, not by the AI.
- Before it goes live, it's tested with your real cases against an agreed exit criterion.
Your data
- We work with models from OpenAI, Anthropic and Google, or with open models that run on your own servers or in your cloud: you decide which one sees your data, and with open models the conversations and documents are processed inside your company, not in a provider's cloud.
- We tell you, with data, when it's worth adapting a model and when connecting it to your documents is enough.
- Every decision that affects a person keeps its reason, and a person can review it. Mexico's 2025 personal data law gives your customers the right to object to automated decisions made without human intervention.
In our September 2026 measurements, a 20-billion-parameter model runs entirely on a single 16 GB graphics card, at about 60 words per second.
Once it's live, we keep it running: adjustments, model changes and monitoring, as a monthly service.
How we use AI ourselves
Before offering it, we use it every day: we develop, test, document and operate our infrastructure with supervised AI agents, under our own control rules. What we offer you is what already works for us.
If you want the same for your technology team —development, infrastructure and security with supervised agents—, we do that too.
What happens after you write to us
- 1
You message us on WhatsApp with your industry, your role and your main system.
- 2
We schedule 45 minutes with a Hábil architect.
- 3
We ask for the process, the systems it touches and the approximate volume. We don't ask for access or for your customers' data.
- 4
You leave with a one-page brief that tells you whether it pays off, what depends on your systems and which risk you must solve before moving forward.
- 5
If it pays off: a project with a KPI, an exit criterion, and scope and cost agreed before we start, including the processing cost with its cap.
We tell you no when the process doesn't have enough volume, when the rule changes every week or when a mistake costs more than it saves.
Write to us on WhatsApp. In 45 minutes we'll tell you whether it pays off, what depends on your systems and which risk you must solve before moving forward.
Feasibility review on WhatsApp
Prefer email? Write to us at hola@habil.mx