Where Copilot and ChatGPT are genuinely good
Both are excellent writers. Give Copilot a paragraph and a tone and it rewrites it well; give ChatGPT a page of notes and it produces a clean summary in seconds. Inside Word and Outlook, Copilot is the fastest way to tidy an email or restructure a section you have already written. If your team answers a handful of short RFIs a year and has the content at hand, that may be all you need.
They are also already licensed in most organisations, which makes them the default starting point. The question is not whether they can write, it is whether they can run a bid.
- Best for
- Short responses, rewrites and summaries of material you already have, by people who own the process themselves.
- Watch out for
- No view of the tender's requirements, no memory of what won, no compliance matrix and no source for a generated sentence. On a consumer plan, confidential inputs may be used for training.
Where they fall short on a public tender
A tender is not a writing task. It is two hundred requirements with weights and knock-out criteria, a deadline, amendments that change the requirements halfway through, and an evaluation panel that scores you on whether each requirement is answered where they expect to find it. A general assistant has no structured view of any of that. It will write a confident paragraph about your service level, but it does not know that requirement 5.1 asks for 99.9% availability and that your standard answer says 99.5%.
It also starts from zero every time. The CVs, the references, the product documentation and the sections that won last year are the raw material of a good bid, and they live in your systems, not in the model. Copilot can read files you point it at, but it does not know which proposal won and which did not, and it will not tell you which page a sentence came from.
What Pentimenti does differently
Pentimenti is built around the tender, not around a chat box. Atlas, the orchestration agent, reads the whole pack, extracts the requirements into a compliance matrix, and spawns the specialist agents the response needs: product, CV, reference, contract, compliance. Each one drafts from your own approved content and attaches its sources. A named person on your team reviews and approves each section; nothing goes out on the agent's say-so.
Because the data is the bid team's institutional memory, it stays yours: GDPR compliant, never used to train models, with an EU-only hosting tier and a single-tenant EU option for the most sensitive bids. And because tenders have to be found before they can be won, TenderScout watches TED, Udbud.dk and Doffin overnight and matches notices to what you actually do.
- Best for
- Teams that answer public tenders and formal RFPs and need every answer traceable, compliant and in the EU.
- Watch out for
- It is a bid platform, not a general assistant. Most customers keep Copilot for email and documents and run the bid itself in Pentimenti.
Using both
This is not an either-or. Copilot stays the assistant in Word, Outlook and Teams; Pentimenti is where the tender lives from the first notice to the submitted proposal. The output of Pentimenti is a Word or PDF document, so the final polish can still happen in the tools your team already uses.
