Evidence Before Rollout: Fulton Hogan's M-Files AI Assessment

The Problem To Be Solved

Fulton Hogan knew it wanted AI applied to document control and QA. What it did not have was a defensible view of which platform to use for what, or what a rollout would cost and risk. Three candidates were on the table with overlapping capabilities: M-Files Aino, Microsoft Copilot, and BlueGPT, the company's own GPT-5-based solution. Open questions ran from Copilot licensing and indirect access pathways to whether M-Files should move to Cloud, and none of them had answers grounded in anything more than vendor material. 

Why Solving The Problem Was Important

Document control and QA are high-volume, high-stakes work at Fulton Hogan. Notice responses take roughly an hour each and happen daily. Metadata tagging was reported at three or more hours per day per person across a user base of 300-plus. Tender feedback had already made AI a competitive question rather than an efficiency one. 

The risk ran the other way too. A poorly governed rollout could expose sensitive project content, break permission trimming, or route people around established processes. For a business where records management is part of how projects get delivered and defended, that is a compliance problem, not an IT inconvenience. 

How TEAM IM Got Involved

Fulton Hogan moves faster on AI than most of its industry does, beginning this project with a working group running and real long-term investment behind it. The open question was how the myriad use cases they were evaluating would intersect with the business context managed in M-Files. We raised that intersection off the back of a long M-Files relationship, and rather than pitch a platform, we proposed a stream of work to ground and prove out the use cases that touched the vault. Scott Olesen, our VP of Services, took delivery of this one himself rather than handing it off, with Dwayne Parkinson and Fletcher Thomas as Solution Architects. The work was scoped under the existing Master Services Agreement and ran across four workstreams over about a month.

The Method

Rather than compare tools on paper, we built in Fulton Hogan's own environment and let observed behavior settle the arguments. 

We started by confirming requirements and evaluation criteria, then configured development and pilot vaults: installing M-Files Aino, connecting M-Files to Copilot through the M-Files Connector, standing up Azure services and Copilot Studio, and syncing representative project data. From there we built proof-of-concept artifacts for the three highest-value use cases. 

Notice summarization and response drafting. A Copilot Studio agent that reads an incoming Notice, summarizes it, and drafts a response, using Power Automate to reach the M-Files REST API. 

PCP and QA review. An agent that reviews Process Control Plans for failed and missing test records, the kind of gap that is expensive to find late. 

Automated metadata suggestion. Aino applied to the metadata properties the workshops identified as priorities. 

Alongside the builds, we scored six candidate use cases against Fulton Hogan's own AI Solution Weighting model, ran workshops with Document Controllers and QA Managers to pressure-test the results, and worked through licensing, indirect access pathways, and BlueGPT integration feasibility. Findings in the report were graded by strength of evidence, separating what we directly observed from what stayed directional. 

The End Result

The delivered Assessment Report set out a multi-platform position rather than a single-vendor recommendation: Aino for vault-native document intelligence, Copilot for enterprise-wide productivity, and BlueGPT reserved for funded, governed custom scenarios. Licensing was modeled as a phased POC, Pilot, and Rollout sequence with decision gates, so spend tracks proven value instead of a projection. 

The clearest signal of how it landed is what happened next. Fulton Hogan committed to a follow-on M-Files AI Pilot Implementation, taking the highest-value use cases into tightly scoped pilots and validating permission trimming and auditability before expanding. 

Conclusion

The harder findings were not about the models. The biggest risks we surfaced were organizational: permission drift, shadow repositories, unclear ownership of content. Those are the things that turn a working demo into a governance incident at scale, and they are the reason the recommended path is gated rather than fast. TEAM IM continues as Fulton Hogan's M-Files partner into the pilot implementation, carrying forward the agents, the governance approach, and the decision gates defined during the assessment.