A general chat assistant drafts individual answers quickly and well. A tender procedure, however, consists mostly of the steps around the writing: finding the publication at all, capturing every requirement completely and by type, tracking the submission and clarification deadlines, evidencing certificates, approving as a team, keeping a record in case the award is challenged, and writing the answers back into the buyer's workbook with its structure intact. Those steps decide exclusion or award. Tendaro uses language models itself and adds exactly that process around them. Basis: August 2026.
You are not choosing between Tendaro and a language model. Tendaro uses language models. The question is not whether the model writes well, but what else has to happen before a response is submitted on time, complete, and in the buyer's own format.
A chat assistant can draft a good answer to requirement 47. It cannot tell you that requirement 47 exists, that it is mandatory, who owns it, that it contradicts what you told this same authority last year, that it is due Thursday, and that it has to end up in cell F47 of the issuer's workbook.
We are not going to argue otherwise. For a single passage, a rewrite, a translation or a first outline, a language model is fast and good. If you answer two or three small tenders a year, hold the documents in your head and write every line yourself anyway, you do not need a platform. In that situation the chat window is the cheaper tool.
The arithmetic changes as volume, formal requirements and multiple contributors arrive. The effort moves away from writing, and that is where Tendaro starts.
Tender documents are third-party documents you received under a confidentiality undertaking from the awarding authority. Two consequences follow, and neither is about how good the model is.
First, pasting text from a third-party PDF into a chat window hands that document's entire contents to the model's instruction layer. Tendaro treats requirement and response text as data, never as instructions.
Second, in a personal chat account each individual decides their own settings, on an account your company does not administer. You cannot answer afterwards which pricing structure went to which provider and when. With Tendaro there is one processing path, tenant-isolated storage in Swiss data centres, an audit log and settings at organisation level. Customer data is not used to train third-party models, contractually excluded rather than toggled off in a personal account.
For requirements that go further there is a second tier. In defence, healthcare and critical infrastructure, or where your own client contracts forbid processing abroad, we run Swiss model environments, so inference does not leave the country either. The question then is not whether we can meet your bar, but where your bar sits.
If your bid work consists of a few small, informal requests, a language model on its own is the cheaper choice, and we will say so plainly. Once you respond to public tenders regularly, submit in the buyer's template, track deadlines and evidence, and involve more than one person, you are not comparing two tools. You are comparing a tool with a process. The model writes. Everything before and after it is the work.
Basis for this comparison: August 2026. Statements about general chat assistants refer to the usual consumer products as publicly documented at that time, without connected specialist services, and not to models accessed through an API. Where a chat client is wired to a tender database through a connector, the line moves: search and a first assessment are then covered, while evidencing completeness, approvals, the audit record and the structure-preserving export are not. Products and contractual terms change; the provider's current documentation governs.
| Step in the procedure | General chat assistant | Tendaro |
|---|---|---|
| Finding the tender | Not covered, you bring the documents | Monitoring across all 26 Swiss cantons, weighted by company profile |
| Capturing requirements | Whatever you paste in | Fully extracted, split into must, should and can, with source location |
| Company knowledge | Re-uploaded every session | Persistent knowledge library, proposals with source and fit score |
| Contradictions in your own material | Invisible | Flagged before both answers are submitted |
| Evidencing completeness | Not possible | Countable list with a status per requirement |
| Deadlines | No clock, no calendar | Submission and clarification deadlines with reminders |
| Submission | Text in a browser, transferred by hand | Answers in the original template, structure and formulas intact |
| Collaboration | One private tab | Roles, assignment by discipline, one shared state |
| Evidence when challenged | A chat transcript | Version history, audit log and staged approvals |
| Data handling | Personal account, per-person settings | Tenant-isolated in Swiss data centres, settings at organisation level |
For a single passage, you can. For a full procedure the steps around the writing are missing: finding the tender, capturing every requirement completely and by type, tracking deadlines and evidence, approving as a team, keeping a record of the process, and writing the answers back into the buyer's own template. Those steps decide exclusion or award, and a chat window covers none of them.
Yes. Tendaro orchestrates language models for extraction, matching and drafting, and adds what a model alone cannot do: access to your company knowledge, a countable requirement structure, deadline logic, approvals, an audit log and a structure-preserving export. Every output is a source-linked suggestion that a person reviews and releases.
Customer content is processed tenant-isolated in Swiss data centres and is not used to train third-party models. Requirement and response text is treated as data and never executed as instructions, which makes instructions embedded in third-party documents inert. Access and changes are logged and controlled at organisation level.
When you answer a handful of small, informal requests a year, read every document yourself anyway, work alone, and do not have to submit in the buyer's template. The chat window is then faster and cheaper. We do not recommend a platform in that situation.
Because many awarding authorities expect the response in their own workbook, with answers in exactly the intended cells. The text takes minutes to write; transferring two hundred answers into someone else's grid takes days and produces exactly the errors that trigger formal objections. Tendaro writes every released answer back into its original cell in the original file.