What an RFP Expert asked me about AI in RFP tools, and how we answered it
A proposal writer who wanted to demo Diligio Respond emailed me last week. They run a team and are shopping for a new AI agent and content library solution to speed up their RFx process. They have used one of our largest competitors (Responsive) for the last couple of years and have been demoing the solutions offered by most of our major competitors. What they shared and what they asked me made me think of writing this post for everyone on how we work.
Their complaint about most existing products was that the tools still require a massive amount of human intervention, and the AI capabilities feel weak for what they need. They wanted to know how we handle that, plus user management, output quality and new feature integrations. Fair questions, I used to manage RFPs on Loopio, I ran into the same problems, and Diligio is basically the tool I wish I had then.
The root of the trouble is age. Loopio and Responsive were originally built on NLP and keyword matching, which still shape their workflows. It shows up as confirmations: confirm to import, confirm to assign, few bulk actions, confirm the answer, confirm again (if you’re sure about the previous confirmation from 2 seconds ago). The interface was designed for a person who searches, not a system that retrieves, which is why the new AI features feel bolted on to a workflow that was never rebuilt around them.
The answers I provided them are given below:
01. User management
Pricing first, because seat-based pricing is what forces proposal teams to ration their own tool. The thing a team actually wants is every subject-matter expert answering directly, and per-seat pricing punishes exactly that. Diligio costs USD 7,499 a year, flat, up to a thousand users (for good measure), every feature and add-on included. No tier where the useful workflow sits behind an upgrade.
The other half is clicks. We designed the workflow in the opposite direction from the incumbents: strip out every unnecessary click and confirmation for admins and non-admins, and keep the ones that carry a decision. For power users, we allow them to work using their own corporate AI agent over MCP or REST API and use tools like Claude Cowork or ChatGPT Work out of the box, no clicks necessary.
02. Output quality
Hallucination is the thing everyone is afraid of, so the architecture deserves specifics. We use two different LLMs, one writer and one verifier, both of which independently rank first in external benchmarks, as well as our 160-question reference internal testing for low hallucination in vector database search. The writer drafts inside a deterministic workflow, constrained to your approved content rather than free to improvise. The verifier checks that draft against the same content, and it is measured on saying "Answer not found", not on pushing an automation score up. A tool that declines to answer is worth more than one that confidently invents your data retention policy.
Every suggested answer carries its specific source in your approved Q&A content, so verifying a claim is one click rather than an act of faith. Using a suggested answer marks it as fresh automatically, which retires the monthly review that our competitors require but nobody completes. Provenance is what makes that survivable.
03. Integrations
All our integrations and features are included in the flat price, and you don’t need to organise a demo to see how much we charge. For our clients, we promise to take any integration requests and fast-track them within two weeks. This is a promise, not a roadmap item. We are a fast and lean start-up; we work that way.
If you are someone who’s evaluating RFx, RFP or DDQ automation solutions for your company, I would love to give you a short demo and answer any questions. And if you are someone who would rather chew glass than talk to a salesperson, the website has a 14-day free trial that only needs a business email address to activate.