Skip to content

Client portals and integrated AI tools

Custom client portal: the space where your clients follow their files

A client portal replaces the email back-and-forth and the phone calls asking where things stand. It only works on one condition: that it shows your clients information coming from the system, not a copy kept up to date by hand.

What separates a useful portal from an abandoned extranet

Many extranets were built and then left to rot, and the cause is nearly always the same: they required someone internally to put information into them. A client space fed by hand is a second round of data entry; it gets filled for a few weeks, then falls behind, then becomes wrong — and a wrong space is worse than no space.

A useful portal duplicates nothing. It opens a window onto the data your teams already produce by working: the state of a file, the documents received, what is missing, what has been invoiced. Nothing is published because someone remembered to publish it; everything is visible because the system already knows where the file stands.

The benefit runs both ways. Clients stop calling to ask where things stand; your teams stop re-sending documents by email that were already sent. And the portal becomes the place where clients supply what they owe you, with validation at capture rather than an attachment to rework.

Where artificial intelligence has a place in this, it belongs inside the tool, on your data, with a source cited on every answer. AI that answers without saying where the answer came from moves the verification work rather than removing it.

What the work covers

File status, live

What the client sees is what the system knows, at the moment they look. No manual publishing, so no gap between the internal screen and the external one.

Document upload with validation

Expected documents are requested by name, checked at capture, and attached to the right file. What is missing is flagged to the client rather than to your team.

Accounts, roles and organisations

One client, several contacts, different permissions; a network head and its members; a principal and its subcontractors. The portal’s structure follows the structure of your commercial relationship.

Contractual documents

Quotes, contracts, invoices, reports: viewable and downloadable from their source, with their history. This is what ends the requests to re-send.

AI inside the tool, with its source

Search your documents, draft a reply, extract a value from an incoming file — where the work happens, and with the exact passage the answer rests on. An answer without a source cannot be acted on.

What the portal does not show

Margin, internal costs, internal discussion: the boundary is decided explicitly and enforced by the system. It is the first question to settle, before any consideration of screens.

The signs a portal would save you time

A portal is justified by the volume of exchanges it removes, not by the impression it gives.

  • Your teams answer the same progress questions every week.
  • Supporting documents arrive by email, as attachments to re-file.
  • A document already sent is regularly asked for again.
  • An extranet exists but nobody keeps it up to date.
  • Your clients or members have no way to consult their own history.
  • Information given verbally gets lost between two contacts.

Frequently asked questions

Do we need an existing ERP to open a client portal?
You need a source of truth, whatever form it takes: a management system, an internal database, sometimes a set of rigorously kept files. What does not work is a portal fed by hand alongside the real work — it falls behind, then becomes wrong. Where no usable source exists, building one is part of the job.
How is client data kept separate?
Separation is enforced server-side, on every request, from the account making it — not by a display filter. A portal whose isolation depends on what the interface chooses to show is a portal where a guessed address opens someone else's file.
Can the embedded AI get things wrong?
Yes, which is why every answer comes with the passage it rests on. The intent is not to decide on someone's behalf but to find information faster while keeping verification possible. On decisions that commit the company, AI prepares and a person approves.
Where is the data hosted?
On infrastructure you can take over, within the European Union, with documented access and backups. It is a point to settle at the start of the project rather than at launch, because it drives technical choices that are hard to undo later.
Can we start small?
It is the best way. A first version that shows the state of a file and accepts a document already removes most of the chasing. The rest — history, invoicing, messaging — is added once you have seen what users actually try to do.

Let’s talk about your project

Describe what you are trying to build. We answer with an honest first read on feasibility and scope — even when the answer is that you do not need us.

Tell us about your situation