Guides ·
AI for accountancy practices without sending client data to the cloud
An accountancy practice can use AI that runs on its own premises, so the models and the memory built from client records sit on an installation in the practice’s building rather than in a provider’s cloud. Seesa is built this way. It helps your team find context, prepare follow-ups and organise recurring work across the client information it is permitted to use. Seesa prepares the work; your team reviews advice, confirms deadlines and makes submissions. It does not file with Companies House or HMRC.
Why client data makes this a serious question
Accountancy practices hold detailed records about their clients: ledgers, correspondence, working papers and documents of every kind. Many of the questions that take time are really searches across those records. What did the client send last quarter? Which documents are still missing? What was agreed about the fee?
AI is well suited to that kind of search and preparation, but it makes the question of where client information goes more immediate. With a cloud AI service, the work happens in the provider’s cloud. With Seesa, the processing and the company memory run on your installation, and you decide who can see what.
What Seesa does on your premises
Seesa’s models, company memory and agent execution run on the installation that Pentatonic supplies, installs and manages. It connects to the places where your records already live, such as email, calendars, documents and conversations, and builds one memory from them. Answers come with the sources behind them, so a person can check where something came from.
That is the foundation. On top of it, the Seesa accountancy page lists the kinds of client work it can help move forward.
- Requests for missing records.
- Preparation for a client review.
- Fee follow-ups, and checks against recorded filing dates.
Your team stays responsible
The line is clear on Seesa’s accountancy page: Seesa can prepare the work, and your team reviews advice, confirms deadlines and makes submissions. A recorded filing date is something Seesa can check against, not something it confirms on your behalf.
In practice that means a draft is a draft. A follow-up email is prepared for a person to read, edit and send. A review pack is assembled for the person conducting the review. Seesa’s own website terms add a general reminder that AI-generated outputs can be wrong and should be reviewed before being relied on. Seesa is there to save the searching and the first draft, not to take the decision.
Keeping client access deliberate
Practices rarely want everyone to see everything. Seesa agrees permissions and connection scope during setup, so the information available to each person matches their responsibilities. Each person’s access governs what their Seesa can use, and sharing an answer does not have to mean sharing the records behind it.
You also set the permitted actions and approval rules for connected tools. Your team decides where Seesa can prepare work, and where it can carry out actions that you have agreed.
What on-premise does and does not mean
It is easy to over-claim here, so it is worth being exact. On-premise does not mean disconnected. Seesa’s security page says source systems, user access and maintenance may use external services. If your records sit in Gmail, Microsoft 365 or Google Drive, Seesa connects to those systems to work with them, and the company explains these connections as part of the deployment design.
What on-premise does mean is that the models, the company memory and the agent work run on your installation, and that company memory is stored there. The practice also gets a documented support connection and data-access arrangements before the system goes live.
Getting started
A practice does not need to start with everything. The accountancy page suggests picking one recurring piece of client work that your team wants to prepare better, and building from there. Standard setup includes installation, historical import, supported connections and permission configuration. Where a routine is particular to your practice, optional specialist engagements can build it with your team, under a separately agreed scope.
Seesa serves companies of all sizes, and the accountancy page shows how it applies to practices. The security page covers access and external connections in more detail.
Routines for the work you repeat
Practices have work that comes round again and again: chasing the same missing records, preparing for the same kind of review, following up fees. Seesa lets regular work become a routine, written in plain English. Seesa’s film says more than a hundred routines are ready to switch on, that you can write your own, and that Seesa learns from the work you do and suggests what to automate next.
A routine has a trigger, a piece of groundwork that Seesa does, and a point at which your team takes over to review, refine and share. That last step is where your professional judgement sits. If you would like help building one that is particular to your practice, the specialist engagements are optional and priced separately from standard setup.
Questions, answered.
Does Seesa file returns with HMRC or Companies House?
No. Seesa’s accountancy page says it does not file with Companies House or HMRC. Your team reviews advice, confirms deadlines and makes submissions.
Is client data stored in the cloud?
Seesa’s company memory is stored on your installation, and the models and agent work run there. On-premise is not the same as disconnected: source systems, user access and maintenance may use external services, which Seesa explains as part of the deployment design.
Can different people see different clients?
Permissions and connection scope are agreed during setup so the information available to each person matches their responsibilities.