private-local.ai vs Seesa: measure the machine first, or lease the finished system

Published October 13, 2026Updated October 13, 202611 minutes read

private-local.ai vs Seesa is a comparison between a finished system and the test that can come before it.

Seesa, made by the London company Pentatonic, installs an artificial intelligence (AI) system in a company's own building. The system reads the company's email, documents and chat, answers questions with cited sources and prepares recurring work for people to review, and Pentatonic manages it. The hardware can be bought or leased, and Seesa's calculator prices a lease of 36 months: for 50 people, £5,887.62 a month, which by our arithmetic comes to about £211,954 over the term. Seesa states that the capacity of the calculator's configuration has not been benchmarked on that exact hardware, and that it verifies speed and concurrency before it proposes hardware.

We offer a step that can come before such a commitment: one or two hours on a named machine and open model, with the tools that you already use and a representative task. Our test drive runs on the NVIDIA DGX Spark, and for a team of 25 people, Seesa's calculator proposes two DGX Spark machines, sized for 3 simultaneous requests, at £89.96 per person a month.

What to establish before signing a lease

Seesa publishes pages for three industries: property management, accountancy practices and insurance brokers. In each, the work it offers to prepare depends on how well an open model finds and uses the right records, and on how quickly it answers when several colleagues ask at once. A test drive shows both for an open model on a DGX Spark. Seesa proposes that machine for small teams and larger servers for larger ones, and it does not name the size or version of its model, so the test drive shows what this class of machine achieves, not the exact system that Seesa would install.

Industry	Work that Seesa offers to prepare	What a test drive shows first
Property management	The system brings together upcoming renewals and certificate dates, maintenance correspondence and arrears follow-ups for each property, with a proposed next step.	Whether an open model on a DGX Spark picks the relevant record out of years of correspondence and drafts a usable next step, and how long that takes.
Accountancy practices	The system prepares requests for missing records, material for client reviews and fee follow-ups, and checks recorded filing dates.	In our own accounting test, every model's workbook contained errors that only a line-by-line check revealed. A test drive shows how an open model performs on work of this kind and how much checking its output needs.
Insurance brokers	The system gathers the history of each renewal and prepares renewal and claims correspondence and premium follow-ups.	A renewal file combines policies, claims and correspondence, so the question is whether the model keeps track of the whole history at an acceptable speed on a DGX Spark.

private-local.ai vs Seesa: the main differences

Criterion	private-local.ai	Seesa
What is provided	You receive temporary access to a named machine, model and serving configuration through a web address and an access key. The configuration is documented, so that a result can be reproduced later on equipment that you own.	Seesa provides a finished application on the customer's premises: a company memory over email, documents and chat with access rights for each person, answers with cited sources, and recurring work prepared for people to approve.
Who operates the system	We operate the machine and the model for the duration of the test drive, and you connect your own tools and run your own workflows. If you then buy a machine, your company operates it.	Pentatonic installs the hardware and the software and manages configuration, updates and maintenance remotely. The customer provides the location, power and network access, and the people who authorise connections and check the first results.
Where the processing takes place	During the test drive, the work runs on the machine that you selected, which we operate on your behalf. A machine that you buy afterwards runs wherever your company places it.	The models and the company memory run on a server in the customer's building. According to Seesa, users reach it through Cloudflare, and the installation keeps connections to Pentatonic for status reports, updates and remote support.
Which models are used	The test drive uses open models that we name and fix to a specific version, selected to fit the hardware being tested. No proprietary model, whether ours or another company's, runs on the setup.	Seesa uses models from the Qwen family of open-weight models and names neither their size nor their version. It states that company data is not used to train them.
How the costs are structured	The test drive is charged at an hourly rate, which is published on our pricing page. Afterwards, you decide whether to buy equipment.	Seesa publishes a planning estimate, not a quotation: £117.75 per person a month for 50 people on a 36-month lease, excluding value added tax. Of this, £80 is the software, whatever the size of the team.
Which organisations it suits	It suits companies with fewer than 5,000 employees that are considering the purchase of a machine and want to see it perform their own work first. Examples are a manufacturer that wants an assistant for its technical documentation, a law or tax firm that plans to process client files on its own hardware, and an automation agency whose clients ask for workflows that run locally.	It suits companies that want a finished assistant over their records without running anything themselves, and that are prepared to buy or lease hardware for it. Seesa publishes pages for property management, accountancy practices and insurance brokers, and works with companies in the United Kingdom and the United States.

What Seesa is

Seesa is the product of Pentatonic Ltd, a London company registered in England and Wales under number 11621602, which by its own account was founded in 2018. Seesa connects to email and calendars in Gmail or Microsoft 365, to Google Drive, Google Sheets and Slack, and assesses other systems case by case. It ships with 140 ready-made routines, and its specialists build further ones in separately priced engagements.

Installation follows four stages: understanding the company's setup, installing the hardware, importing the history and permissions into the company memory, and putting the system to work. Seesa aims to have it running within a week of the hardware arriving, which its own guide calls a typical figure, not a promise. The hardware can be bought, leased over 24, 36 or 48 months, or replaced by suitable hardware of the customer's own after an assessment, although the calculator prices only the 36-month lease.

For 50 people, the calculator proposes a server with two NVIDIA RTX PRO 6000 graphics cards with 96 gigabytes of memory each, a planning allowance of £60,000 for the hardware, sized for 5 simultaneous requests. The lease charge assumes that 80 per cent of that allowance is financed at an illustrative 9 per cent a year, from a facility that Seesa states is not yet agreed, that the remainder is recovered over the 36 months, and that a markup of 1.5 per cent applies.

Seesa's security page is unusually specific about what leaves the building. According to it, the encryption from the browser ends at Cloudflare; the installation reports its status to Pentatonic every five minutes, without company data, and downloads updates from Pentatonic's registry without uploading anything; and named Pentatonic engineers have remote access, with their console sessions recorded. Seesa states that the status reports can be turned off on request and that updates can be held for the customer's approval. Its own guide states that on-premises does not mean disconnected. Pentatonic holds a System and Organization Controls (SOC) 2 Type II report covering the company, which it provides on request.

Seesa's strengths

It is a finished application. A company receives a working assistant over its own records, installed and maintained, without choosing or assembling any software.
Its prices are published as a worked example, with the hardware allowance, the financing assumptions and the markup shown, which is rare for AI installed on a company's premises.
Its security page names the connections that leave the building, who can access the installation and how that access is recorded.

Seesa's limitations

The commitment is large and comes before any measurement of the customer's own work: about £211,954 over 36 months in Seesa's example. The calculator's configuration has not been benchmarked on that exact hardware, and Seesa verifies speed and concurrency itself before it makes a proposal.
The models are named only as a family, without size or version, so the quality of the answers cannot be compared with published results before a proposal.
Seesa names no customers and publishes no case studies. It states that it is used by companies from start-ups to members of the FTSE 100 index in the United Kingdom and the United States, and we found no independent coverage that confirms this.

Which option suits which organisation

Seesa is the better choice if

You want a finished assistant over your email, documents and chat, installed and maintained by someone else, and you do not want to choose models, software or hardware yourself.
Your work resembles what Seesa offers, such as preparing renewals, follow-ups and client reviews from your own records, and its standard connections cover your systems.
A purchase or a lease of two years or more is acceptable to you, and you are prepared to rely on the vendor's sizing of the hardware.

A test drive first is the better choice if

You want to know what a machine and an open model achieve with work of your kind before you commit to hardware, whether you then buy a machine, lease one or decide against both.
You want to compare open models from different families, such as Qwen, DeepSeek and GLM, rather than accept one family whose size and version are not named.
You want to keep the tools that you already use, provided that they connect to an interface compatible with OpenAI's, rather than adopt a new application.

Sources and notes

All information about Seesa on this page was read from its own website on 13 October 2026: the home, product, on-premises, security, integrations, industry, pricing and terms pages, and its guides dated 5 October 2026. The prices are Seesa's planning estimate in pounds sterling excluding value added tax, with a price basis of 30 September 2026; the figures for 25 people were read after entering that team size in its calculator. The total over the lease is our arithmetic.

The company details come from Seesa's website and from the public register of Companies House. We found no independent coverage of Seesa, such as a press article, a review or a customer case study.

We have neither used nor measured Seesa. If any information on this page is out of date, please let us know and we will correct it.

Further reading: the other four comparisons cover ChatGPT, Palantir, the open-source engine LocalAI and the redaction company Limina, and the pricing page states the hourly rate.

Questions about private-local.ai and Seesa

private-local.ai vs Seesa: which one should we choose?

They answer different questions. Seesa delivers a finished assistant over a company's email and documents, installed on its premises and managed by Pentatonic, on hardware that the company buys or leases. The test drive answers whether a particular machine and open model can do your work before you buy anything. The two do not exclude each other: a company considering Seesa can first observe what an open model on a DGX Spark achieves with a representative task, and then decide with that result in hand.

How much does Seesa cost?

Seesa publishes a planning estimate rather than a price list. For 50 people on a 36-month lease, it shows £5,887.62 a month for the company, excluding value added tax: £80 per person for the software and £37.75 per person for the hardware. Over the whole lease, that comes to about £211,954 by our arithmetic. Seesa states that the number of people, the workload, the data volume and the final hardware cost change the figures.

Which models and hardware does Seesa use?

Seesa states that its models come from the Qwen family of open-weight models, without naming a size or version. The hardware depends on the team: two DGX Spark machines for a team of 25 people, and a server with two NVIDIA RTX PRO 6000 graphics cards for 50 people. For both, Seesa states that the capacity of the calculator's configuration has not been benchmarked on that exact hardware, and that it verifies speed and concurrency before it proposes hardware.

Does Seesa offer a trial before the lease?

Its website describes no trial or test period. The route it describes runs from a conversation and a demonstration to a proposal and a signed agreement, and its terms state that a lease is subject to financing. Our test drive is charged by the hour and involves no lease.

Measure the machine before you sign for one

Choose a setup, connect the tool that you already use and run a representative task. Within one or two hours, you will know what the machine and the model achieve with work of your kind, before any lease or purchase.

Request an endpoint