What AI Proposal Tools Actually Promise
AI proposal tools for architecture firms promise real productivity gains— 20–50% reduction in proposal preparation time, faster RFP analysis, and consistent credential matching7. The promise is not wrong. It's conditional.
The major platforms in the AEC space each take a slightly different angle:
- Kantiv (formerly Joist AI) — AEC-specific "pursuit intelligence platform" for proposal automation, rebranded in 20249
- Responsive (formerly RFPIO) — general RFP automation platform with broad industry application
- Workorb — AEC proposal AI that positions itself as working with real-world, messy data2
- AutoRFP.ai — AI-assisted proposal generation for proposal teams
Every one of these platforms (whether purpose-built for AEC or adapted from general RFP tools) runs on the same fuel: organized, accessible institutional knowledge. Without it, they all underperform in exactly the same way.
According to U.S. research from the American Institute of Architects (March 2025), only 8% of architecture firms have fully implemented AI solutions, with an additional 20% currently implementing1. The market is still early. And early adopters are learning that the advertised results— the draft generation, the credential matching, the compliance matrices— only materialize when a specific prerequisite is in place.
When implementing AI tools for proposal work, every one of these platforms depends on the same thing: organized, accessible, current institutional knowledge. For a deeper look at how engineering and architecture firms are using AI for proposals, the pattern is consistent— firms that see results built their content foundation first.
As Kantiv notes3: "Proposal content isn't a file. It's an argument." The argument only lands when the evidence is findable.
The First Link That Always Breaks
The bottleneck in AEC proposal development isn't content generation— it's content retrieval. Architecture firms aren't struggling to write. They're struggling to find what they've already written.
According to Kantiv's data3, proposal teams spend 25–40% of their effort on content retrieval and adaptation, not on writing. That's a meaningful share of the $12,000–$15,000 that Responsive estimates each proposal response costs to produce6. And most of that retrieval time produces nothing useful— just confirmation that the answer exists somewhere, if only it could be found.
AEC firms' institutional knowledge is scattered across:
- SharePoint folders nobody has trusted since the last migration
- Egnyte or local shared drives with inconsistent naming conventions
- Deltek project files that were never connected to the proposal workflow
- Old email threads where a principal approved the final language
- PDFs of past submittals with no searchable metadata
- A partner's laptop, which is not a content management system
According to Kantiv3, the average ENR Top 500 firm maintains active descriptions for 50–200 projects at any given time. With 3–5 content variants per project, that's 150–1,000 content fragments distributed across shared drives. And that's for the large firms with dedicated BD coordinators.
AEC+Tech names the risk plainly4:
"Garbage in, garbage out isn't just a saying— it's a prediction of project failure when AI amplifies bad data."
The dangerous version isn't a bad draft. It's a confident bad draft. As AEC+Tech continues4: "Feed AI a folder structure nobody has trusted since 2014 and it will confidently cite the wrong precedent on a live proposal— and a confident-wrong answer is more dangerous than no answer at all."
The 25–40% of effort going to retrieval is the pennies. The actual proposal argument— the win themes, the strategic differentiation— is where the dollars are. But the dollars are what firms came for.
Why Smaller and Collaborative Firms Have It Worse
Enterprise AEC firms at least have dedicated BD coordinators and structured asset management systems. Smaller practices— and network-based firms built on deliberate collaboration across offices— are managing the same problem with far less infrastructure.
Link Lab Architecture10— a Belgian architecture and urban design network founded in 2018 by Dennis Delvael, Kevin Huysentruyt, and Steven Vanwildemeersch, with offices in Kortrijk (Belgium) and Lille (France)— represents exactly the kind of practice where this gets hard. Built on an "Open, Collaborative and Curious" philosophy that deliberately breaks from classical hierarchy, the firm's decentralized, multi-office structure is an asset in day-to-day practice. In proposal mode, it becomes a liability.
Partners and project leads across offices maintain their own files, workflows, and reference materials. No single repository holds it all. No one owns the content library, because the content library doesn't really exist yet as a library.
Most vendor data targets ENR Top 500 firms. But a 5–15 person studio arguably has the worst version of this problem: fewer people to maintain content, higher personalization required per proposal, and more institutional knowledge living entirely in partners' heads. You can't read the label from inside the bottle— when you're close to your own work every day, the scattered-content problem is invisible until an AI tool surfaces all of it at once.
The AIA's U.S. research confirms the adoption gap1: large firms drive AI implementation at significantly higher rates than smaller practices. That gap isn't only about budget. It's about foundation.
Three Steps to Fix the Foundation
Before any AI proposal tool can work, three things have to be true about a firm's institutional knowledge: it has to exist in one place, it has to be searchable, and it has to be current. That's the full prerequisite— and it comes before any tool selection decision5.
Step 1: Audit what exists.
Before organizing, know what you have. Run an inventory: past proposals (PDF, DOCX, whatever format), project descriptions, staff bios, credentials, client references, awards. Note what's current versus stale. Note where each type lives— and how far it is from a single shared location. This is the kind of task that takes a day to start and a week to finish properly. Do it anyway.
Step 2: Centralize to one source of truth.
Pick one system and move everything there. For most small firms, that's SharePoint or Google Drive to start. Deltek or Egnyte if there's already existing investment. The specific platform matters less than the discipline of using ONE location. Unanet's guidance on this is direct11: firms must "centralize and clean up their data before they can truly leverage AI"— without reliable information, even sophisticated AI won't deliver meaningful insights.
Step 3: Tag and standardize.
Content is only retrievable if it's labeled consistently. Implement a simple tagging schema: project type, building type, client sector, year, geography, team lead. Project descriptions need version control— date-stamped, with the current version flagged clearly. Bios need a standard format and a designated update owner.
Some of this work can be done with AI— document extraction, classification, metadata generation— but the schema decisions are human. This is the strategy layer. The sequence matters5: content library ready → choose and onboard the tool → see ROI. Reversing this is exactly why rollouts disappoint.
As OpenAsset notes on what always remains human work8: "Win themes should be hyper personalized… AI just can't do that for you."
The hidden costs in AI implementations often show up right here— not in the subscription fee, but in the hours spent organizing content that should have been organized before the tool was purchased.
Choosing the Right Tool — Once You're Ready
With an organized content library in place, the tool question has a clear answer: match the platform to the firm's size, budget, and how many proposals it runs annually.
| Platform | Best For | Requires | Caution |
|---|---|---|---|
| Kantiv (formerly Joist AI) | Mid-size AEC firms with active pursuit pipelines | Deltek, SharePoint, or Egnyte integration; organized content9 | Higher onboarding investment |
| Responsive (formerly RFPIO) | Cross-industry firms with AEC application | Manual content upkeep, rigid library structures5 | Legacy AI layer; heavy human review |
| Workorb | AEC firms willing to pilot newer platforms | Content accessible in some form2 | Newer platform, still maturing |
| ChatGPT + clean library | Small firms, < 20 people, < 50 proposals/year | Organized shared document library7 | Manual process; validates the workflow before specialization |
For small firms starting out, the path is clear. Begin with ChatGPT ($20/month) paired with a clean, centralized document library. Build the habit of content organization and prompt discipline before investing in a specialized AEC platform— purpose-built tools represent a meaningful cost jump and longer onboarding commitment than the ChatGPT starting point. It's the cheapest way to validate whether AI proposal acceleration fits the workflow.
The right platform for a 10-person studio running 30 proposals a year is almost never the same tool built for an ENR Top 500 firm— and the pricing gap between them usually makes the match clear.
What to ask any vendor before signing: "What does onboarding require from our team? What format does content need to be in? What happens when our content library has conflicting versions?" If the sales process doesn't answer these clearly, that's a signal.
And note the market context: Joist AI's 2024 rebrand to Kantiv— same team, "bigger vision"— signals that platforms in this space are still finding their positioning9. Evaluate with a pilot, not a full commitment.
For a structured way to compare options, the AI decision framework for founders lays out evaluation criteria by firm size and workflow. Team adoption follows content organization— once the library is in place, the onboarding conversation with your team becomes much simpler.
What AI Can Do When the Foundation Is Solid
Architecture firms that fix their content foundation first report real, measurable proposal improvements— and they access them in weeks, not months. That's a fast return on what can feel like grinding organizational work.
And with a centralized, current content library in place, these capabilities become available:
- Automated RFP analysis — section-by-section breakdown in minutes, not hours
- Credential matching — the most relevant project descriptions surface without hunting across drives
- First-draft generation — from actual historical content, not generic boilerplate
- Compliance matrix generation — systematic review of requirements against firm capabilities
Firms that have properly implemented AI for proposals report 20–50% reductions in proposal preparation time7. OpenAsset's research puts individual time savings at 5+ hours weekly for marketing teams that have done the setup work correctly8.
The human-AI split is what makes the investment worth it. AI handles the retrieval, the structuring, and the first draft. Humans provide the win themes, the strategic differentiation, the client relationships. That's the architecture of a proposal engine— domain expertise plus organized access to it produces the amplification AI actually delivers.
If auditing the content library and mapping the right tool to your workflow feels like a project that keeps slipping down the priority list, that's exactly the kind of engagement Dan Cumberland Labs runs. That's what our AI implementation work is built around— from content audit through tool selection and adoption.
FAQ
What is the biggest reason AI proposal tools fail in architecture firms?
The content library— not the AI itself. When institutional knowledge is scattered across shared drives, email inboxes, old proposals, and partners' heads, the AI amplifies the disorganization. Organized content in produces organized proposals out4. The tool is only as good as what it has to work with.
What do I need before adopting an AI proposal tool?
A centralized, current, searchable content library: historical proposals in accessible formats, tagged project descriptions, up-to-date staff credentials, and a single location the whole team actually maintains. Without this, even purpose-built platforms like Kantiv or Responsive will underperform5. Audit first, centralize second, tag third— and only then evaluate which tool fits your volume and workflow.
What is Kantiv (formerly Joist AI)?
Kantiv is an AEC-specific "pursuit intelligence platform" for proposal automation— rebranded from Joist AI in 2024 with the same team and a broader product vision9. The platform integrates with Deltek, SharePoint, and Egnyte and requires organized institutional knowledge to deliver its core value. The rebrand signals market ambition; the underlying requirement (clean, accessible content) hasn't changed.
What is "pursuit intelligence" in AEC?
Pursuit intelligence is the practice of using AI to analyze a firm's historical data— past proposals, win/loss history, project performance— to improve go/no-go decisions and proposal quality. Kantiv is the primary platform using this framing9. In plain terms: it's pattern-matching at scale so the firm can chase the right work and build better proposals for it. It's not a substitute for content organization— it's the layer you build on top of it.
How many architecture firms currently use AI?
According to U.S. research from the American Institute of Architects (March 2025), 8% of architecture firms have fully implemented AI solutions, with an additional 20% currently implementing1. Large firms adopt at significantly higher rates than smaller practices. The gap between "using AI tools" (53% of A&E firms) and "integrated into actual workflows" (27%) tells the full story7: adoption is running well ahead of real implementation.
References
- American Institute of Architects (AIA), "New Research Explores Perceptions and Opportunities: Artificial Intelligence" (March 2025) — https://www.aia.org/about-aia/press/new-research-explores-perceptions-and-opportunities-artificial-intelligence
- Workorb, "Your Firm's Memory, Working for You: Knowledge Management for AEC Proposals" (2025) — https://www.workorb.com/blog/workorb-knowledge-management-aec-proposals
- Kantiv (formerly Joist AI), "Why 80% of AEC Proposal Content Goes Unused" (2025) — https://www.kantiv.com/post/aec-proposal-content-black-hole
- AEC+Tech, "A Practical, Purposeful, and Private Approach to AI Adoption in AEC" (2025) — https://www.aecplustech.com/blog/practical-purposeful-private-approach-ai-adoption-aec
- Flowcase, "Best Proposal Automation AI Tools: Features & Comparison (2026)" (2026) — https://www.flowcase.com/blog/best-proposal-automation-ai-tools-features-and-comparison
- Responsive, "The Architect RFP: How to Issue and Respond" (2024) — https://www.responsive.io/blog/architect-rfp
- Dan Cumberland Labs, "How Engineering Firms Are Using AI to Write Better Proposals" (2025) — https://dancumberlandlabs.com/blog/ai-engineering-proposals/
- OpenAsset, "The AEC Marketer's Guide to Using AI for Proposal Success" (2025) — https://openasset.com/blog/ai-proposal-guide/
- Kantiv, "Pursuit Intelligence Platform for AEC Firms" (2025) — https://www.kantiv.com/
- Link Lab Architecture, "Linking People and Spaces" (2025) — https://www.linklab.eu/en
- Unanet, "How AI is Reshaping AEC Business Development and Marketing" (2025) — https://unanet.com/blog/how-ai-is-reshaping-aec-business-development-and-marketing
