Most project management systems (PMS) have added AI features in the last couple of years: a chat box, a search bar, a few "agents." Since your PMS already holds much of your project data, it's tempting to assume that's enough.
In reality, PMS-embedded AI usually falls short of adoption at scale for five reasons: the models underneath are rarely trained on construction data, the pricing model is unpredictable, training and change management get billed as expensive add-ons, project documents were never preprocessed into anything resembling a project knowledge graph, and don’t handle construction drawings well.
A hard hat on a generic AI model
Most AI features inside a PMS are general-purpose models, like ChatGPT, Claude, or Gemini, with a cost markup: the same technology available to any industry, pointed at your construction documents. These models weren't trained on drawings, specs, submittals, and RFIs as a connected system, because construction data is proprietary and largely absent from what's publicly available on the internet. As a result, they struggle the moment a question or task requires connecting multiple documents, processes, and people the way an experienced project manager or superintendent would.
Pricing and enablement costs work against you
Many of these features run on a credit-based consumption model: you're charged for the text and documents you put in and the output the AI generates. Some even charge you daily credits merely to store your data in their AI infrastructure before any work is actually done. It's nearly impossible to predict how many credits a question or task will take, which makes billing unpredictable, a problem for a business that needs to know its costs for the year and its project budgets well before construction begins. That unpredictability discourages your team from using the tool for fear of hitting a limit, and it gives the vendor no reason to make the tool more efficient, since the more credits you burn, the more they earn.
Rollout usually costs extra, too. Training and change management are often sold as a separate statement of work, so the vendor gets paid whether or not the field ever adopts the tool.
No preprocessing, no real knowledge graph
An AI chat window that finds a spec section isn't the same as a system that understands how that section relates to the RFI that amended it, the submittal meant to satisfy it, and the bulletin showing where it is to be installed. That requires preprocessing: structuring unstructured data across project documents, extracting the relevant information, and mapping each data point to every other relevant one before anyone uses AI. In this scenario, a door tag is already linked to its schedule, spec section, and any open RFIs. Skip that step and the model has to re-extract everything at every query, which drives up both inaccuracy and cost. A PMS running a generic model with no additional training can't build that knowledge graph, so its AI can’t accurately execute construction-specific workflows it was never built for.
Failing to meet construction teams where they are
Construction teams are busy and construction documentation is excessive. AI that only works when someone remembers to prompt it won't get adopted at scale. A better model triggers automatically: a new submittal, RFI response, or bulletin landing in the project management system should kick off an AI-powered review on its own, so the team starts from a thorough first pass instead of a blank chat window.
Generic AI also has known limitations with construction drawings, arguably the most critical documents on a project. Unless AI can read the drawing itself, not just the text in a title block or a markup, it's inherently limited for teams who live in the drawings. In Trunk Tools platform data, roughly 10% to 20% of meaningful drawing revisions go unclouded by the architect, meaning a text-only tool will never catch them.
How Trunk Tools solves these challenges
Trunk Tools combines the best of foundational AI models with our own proprietary models, built specifically for construction data and workflows since 2021, using real project data from hundreds of projects with our customers' explicit permission. That training data is always deidentified and aggregated, so nothing proprietary leaves a project record.
Before a project begins, we ingest and structure data into a project "brain" we call Cortex, so agents and queries produce reliable answers instead of best guesses. Trunk Tools prices with one predictable, flat annual rate for unlimited usage and users, so our incentive is to make the tools valuable enough that customers adopt them portfolio-wide, not to meter every question. Training and change management are part of that same commitment: our customer success managers onboard your team, train them, and meet with them as often as it takes to drive real adoption. We even come to your jobsites to deliver training in person at no additional cost.
An AI feature bolted onto your project management system will always be limited by a model that was never trained on construction data or workflows. Trunk Tools was built to unlock what AI can do specifically for your projects.
Last updated August 21, 2026.