Who keeps your RFP library current?

A hand-kept RFP library is two jobs: the folders of past bids, and a separate library that someone has to fill, tag and check while live bids compete for the same hours. In SEQUESTO's agentic operating system (aOS), the Knowledge Hub is that library, and its upkeep happens inside each bid. Content can be read where it already sits, and each item carries an owner and an expiry date, so expired material is flagged before it reaches a draft. A finished response returns its answers only once an owner approves them.

What changes is where the filing work sits, not who decides. Your team still approves what counts as approved content. The final wording of every response stays with the people who sign it off.

That gives you a practical way to judge it. Put one of your own closed bids through the Knowledge Hub and through your current library, and check two things: whether out-of-date items are flagged before they reach a draft, and whether every match shows the source it came from.

How one question moves through the library

Content comes in from wherever it sits, with its owner and expiry date. It is matched by meaning with its source shown, drafted in the tender's language, and returned to the library only once its owner approves it.

The example that follows is illustrative, not a customer case. A German tender asks how you protect personal data during service delivery, and it asks for your current information security certificate as evidence. Your approved answer, the data protection policy behind it and the certificate all sit in English. Last year's certificate is still in the library beside this year's.

Bringing existing content in

Existing content goes into the Knowledge Hub as it is. Word, Excel, PDF, slide decks and images are converted and indexed automatically, scans are read with OCR, and the files can be in any language.

On the Scale and Enterprise plans, SharePoint and Google Drive feed the Knowledge Hub from your existing repositories without moving the files. The SharePoint integration page shows what the connection reads and what stays in the source system. Before you connect anything, ask where a connected file's owner and expiry date are set. That tells you whether the governance lives in SEQUESTO or in your repository.

What each item carries

Each item keeps an owner, a review cycle, an expiry date and its version history. Any earlier version can be restored. The original upload can be opened beside the converted text, which matters when a reviewer has to check a converted table against the PDF it came from.

In the example, last year's certificate has passed its expiry date, so it is flagged before it can reach the draft. This year's certificate is the one the answer uses. The wider properties of an auditable answer library are covered separately. The point here is that expiry acts before drafting, not at the final read.

Finding content for the question

The Knowledge Hub finds content by meaning rather than by keyword, once James has decided what the question needs: the approved data protection answer, the current policy and the certificate. Each match comes back with a confidence score and the source snippet it came from.

If the bid covers one service line, the team can limit the search to that part of the Knowledge Hub. A case study from another line then stays out of the draft. Agent Force drafts the answer at question level from that approved content and the project's context, and it shows the reasoning behind the draft.

Answering in the tender's language

A team can answer a tender in one language from a library in another. SEQUESTO works in more than 30 languages, so the German response draws on the English content without the team keeping a German copy of the library. The response is exported in the language the submission requires.

Someone who reads German still reviews the wording before it reaches the bid manager. For keeping one process across languages, see multilingual bid management.

Returning the approved answer

When the response is finished, its approved answers flow back into the Knowledge Hub. They don't count as approved content when they arrive, because the owner approves each one first. In the example, the data protection owner reads the German answer and decides whether it joins the library. If it does, the next German tender starts from it.

Seeing who owns what

A dashboard shows the health of the Knowledge Hub: who owns which items, and what is expired or due for review. Each person also has a personal dashboard of everything assigned to them.

What the library does not do

The Knowledge Hub approves nothing. It flags, matches and suggests. People decide what counts as approved content, what is written and what is submitted.

An expiry flag is only as good as the expiry date behind it.

Neither the Knowledge Hub nor James writes the final wording. Agent Force drafts, and James keeps to the project's rules and asks before any action outside the agreed scope. Your team accepts, adjusts or rejects what they produce. How far a reused answer needs tailoring for this buyer stays the writer's judgement. When the response is ready, a person on the team submits it in the buyer's portal. SEQUESTO doesn't submit on your behalf.

A weak match tells you something useful. Each match comes with a confidence score, so the team sees how well it fits before using it. Where nothing fits, the gap is for a person to fill, not the nearest text the library can find.

The library makes no promise about the accuracy of what is drafted from it, and that is why approval stays with people. In McKinsey's 2025 global survey on the state of AI, nearly one-third of all respondents reported consequences from AI inaccuracy. It was the risk they had most often experienced.

On the Team plan, existing content comes in by upload rather than through SharePoint or Google Drive.

Where the library sits in the bid process

The Knowledge Hub is the Context & Knowledge layer, the fourth of the eight layers in aOS, and it works between three roles. James chooses what each question needs from the library. Agent Force drafts each answer from approved knowledge. The bid manager gives the final approval on what goes out.

The library is not the whole project. Each project keeps its own files, deadlines, contributors and working material. Its details, such as the buyer, the scope and the win themes, give James the context of the bid. Connected systems supply product, commercial and compliance context without being copied into the project.

Access follows your structure. Permissions are set by user and role, down to file and folder level, and they are enforced in the data layer as well as the interface. Content doesn't cross project boundaries, and an admin can check the access any person or team actually has on a file. Every retrieval is logged with actor, timestamp and item, alongside edits, approvals and exports. The log can be searched and exported for a compliance review or a buyer's due diligence questionnaire.

In our view, the library carries the three things SEQUESTO is built on, in this order. Governance comes first: the whole response process runs on governed workflows shaped to your own processes. Human-in-control follows, with James orchestrating and Agent Force carrying out the work, while the bid manager owns strategy, tone and the final word. Auditability completes it, because every answer drawn from the library is cited to its source and traceable, so the team and legal can stand behind it.

That fits a broader finding in McKinsey's survey, although the survey doesn't look at bid work. High performers there were more likely to have defined processes for deciding when model outputs need human validation. For how the library feeds the rest of the response, from receiving the documents to the final pack, see RFP response.

Test it against your current library on a closed bid

Run one closed bid through both libraries, with the same person answering, and judge each on the same marks:

• seeded stale items flagged before they reach a draft
• every match sourced
• the tender's instructions met
• the people-hours each role spent

Unlike a demo, which the vendor steers, this test runs on your content, with your own staff at the keyboard, against traps the vendor hasn't seen.

What to measure

Choose a closed bid whose questions cover your recurring sections and some evidence-heavy ones. Fix the question set before anyone starts.

Then ask someone outside the test, such as a content owner, to seed both libraries alike with known stale items. Good candidates are an expired certificate and a superseded policy version. Only that person keeps the list.

The person answering alternates which library goes first, question by question. Answering a question the second time is usually easier, and alternating spreads that advantage evenly between the two libraries.

Judge each library on these marks:

• Stale items, judged per instance. Any seeded item that reaches a draft without a flag fails that library. The limit is zero.
• Sources, judged per question. Every match shows the file it came from, and the reviewer checks it by opening that file. One unsourced match fails the question.
• Precision, judged per question against your current library. Word and character limits and the tender's own instructions are met. The new option passes with no more breaches than the current library produced.
• Quality, judged once. A reviewer who doesn't know which library produced which answer scores both sets against the tender's award criteria. The new option passes if it scores no lower than the current one.
• Time, recorded per question. Record people-hours by role: the bid manager, the person answering, subject matter experts, and anyone in legal or compliance who reviews.

What the test costs

The test costs people's time as well as licences. Count the hours of whoever seeds the stale items, the person answering and the blind reviewer. Count the hours of the owners, subject matter experts and legal reviewers who confirm items and approve what returns. Don't count only the bid manager's time.

Three costs repeat for each option:

• Set-up. SEQUESTO's free trial on the Team and Scale plans needs no credit card. It opens with an intake call with a solution engineer and a one-hour onboarding session, followed by the content upload. Your current library needs preparation too, since the seeded items go into both.
• Clearance. InfoSec and whoever governs the closed bid's documents must agree before those files go into a trial environment. Your current library is already cleared, which is a real point in its favour.
• Training. The person answering needs enough time in the new tool to give it a fair run. Record those hours apart from the per-question times, so the comparison measures answering, not learning.

If you want a first look before committing staff, SEQUESTO runs demos on a closed tender you bring. A demo is a useful preview, but the seeded run needs your own person driving it.

Keeping or moving

Decide in two steps, and count each gain once. First, an option fails if it misses the stale-item, source or precision marks, no matter how much time it saves. Second, among options that pass, the quality score is only a floor, and time decides. Time here means total people-hours across roles for the run, plus the cost of switching. That way, time saved isn't counted twice as a better score.

Size the switch honestly. Every plan includes onboarding that covers data upload, configuration and team training. In our own observation, not a published benchmark, teams with an approved answer set are usually productive within two to four weeks, and teams starting from scattered documents take two to three months. On Scale or Enterprise, connecting SharePoint or Google Drive means the files don't have to move, though each item still needs an owner and an expiry date wherever those are set.

Keeping your current library is the right result if it passes the same marks and your process is simple and stable. In that case, switching buys little. If your current library fails the stale-item test, you have learned something worth acting on whichever way you decide.

This week, pick the closed bid and the person who will seed the stale items.