AI Platform Engineering
AI inside the portal your customers already sign in to.
We engineer AI into the portal your customers already sign in to: an assistant that answers from their own policy, plan or account, forms that read the documents they upload, claims that arrive summarised, and a handover to your team when a person should decide.
Each feature is wired into the policy, billing, loan or CRM systems you already run, sees only the signed-in customer's records, and is tested against your own cases before a customer sees it.
why did my bill go up this month?
This month's bill, explained
Heating use rose in the cold spell
Introductory discount
Ended on your last bill
Meter reading
The last one was estimated
Plans for how you use energy
Compared on your own usage
A customer's question is answered from their own bill, plan and readings, with the lines that explain it on top.
Answers from the customer's own policy, plan or account, and hands over to your team when it should.
Hit by a reversing car while parked: rear bumper and light damaged. No injuries, no photo yet.
First notice is gathered, checked against the policy and summarised, then waits for a person before it moves.
Assistant, form and claims changes are checked against your own past cases before customers see them, and the last good release is one step back.
The pieces we build most often. Each one connects to your policy, billing, loan or CRM systems, sees only what the signed-in customer may see, and is judged on your own data.
An assistant in the portal, the app and your messaging channels that answers from the customer's own policy, plan, bills and history, shows the clause or line it used, and acts only within limits you set.
Forms that ask only what applies to this customer and prefill what you already hold. Uploaded payslips, IDs, bills and photos are read into the right fields, and anything the model is unsure of goes to a person.
First notice of loss by text, voice or photo. The portal asks the follow-up questions this policy needs, gathers the evidence, checks cover and hands the adjuster a structured summary to decide on.
Plain-language questions that check eligibility against your product and credit rules and return a quote the customer can follow, with every rule that applied visible to your underwriting or credit team.
Bills, usage and statements explained from the account's own data: why the total changed, which charges are one-off, and which plan suits the way this customer actually uses the service.
Renewal, payment, document and appointment reminders timed to each customer's history, written in their language, sent on the channels they chose, and stopped as soon as the task is done.
When a person should take over, the conversation, documents and account context reach your team as one summary, so the customer never repeats themselves and the person picking it up starts from the facts.
The same assistant by text or voice, in the languages your customers speak, tested on your own call recordings and transcripts, with a person one step away on every channel.
The same patterns serve shipment tracking and patient portals: delivery questions answered from live tracking data, and appointment, referral and billing questions answered only from records the patient is allowed to see.
Agentic access
Customers are starting to hand routine tasks to AI assistants: checking a balance, comparing a renewal, sending a meter reading, chasing a claim. Those assistants need a way into your portal that is safer than screen-scraping a sign-in page.
We build that way in: an MCP server over the account actions you choose, verified agent identity, the customer's consent recorded for each action, and limits on what an agent may read, change or pay. Anything with money or cover attached waits for the customer to confirm.
Most portals start with one journey, such as a bill question, a claim or an application. We build it inside your stack, prove it on your cases and hand it to your team.
A short readiness sprint in your portal analytics, contact reasons and case data to choose the first journey and agree how it will be judged.
Engineers work in your repositories and your cloud, connecting the journey to the policy, billing, loan or CRM systems behind the portal, with access scoped to each customer's own records.
We test on your own past cases before launch, then release to a share of customers and watch completion, handovers and errors, with a rollback ready.
Runbooks, dashboards and a trained team. You own the code, the prompts and the evaluation set; we can stay on to run it if you want.
Tell us what your customers come to the portal to do and where they give up, and we will come back with a first view of how we would build it.