From Skepticism to Structure: How a Scaffold & Access Services Company Found Safe Value in AI

A case study from a Scaffoloding & Access Services company.

A scaffolding and access company, skeptical of AI, asked us what 'Simple AI could they adopt'?

Whilst they were skeptical of both AI unknowns and adding complexity they didn't undertand, they acknowledged their staff were using it already, and receiving increasing requests to provide tools for staff to help reduce their workloads, especially the adminsitrative and complance aspects aspects which were growing.

Leaders were wary, but supervisors, admin, and site managers were already using ChatGPT style tools for scaffold design notes, safety paperwork, hazard prevention, and certification tasks.

The problem?
They were doing this work in good faith, with personal and free accounts, but there was no record keeping or traceability - a serious risk in safety-critical work.

Step 1 – Finding the truth

We ran an AI Readiness Audit to map out risks and opportunities.
We found:

• Strong demand from staff for help with time-consuming, error-prone safety and compliance documents.

• AI already in use, but without controls.

• Leadership, firstly unaware of the extent of usage, and during the audit, increasing concerned about legal exposure, wanting a cautious, governed approach, if they were in fact prepared to adopt AI at all.

• It remained clear at this point, leadership had reservations about any significant AI adoption, but recognised their staff's desire to reduce workloads, and commended their initiatives to find solutions.

The proposal.

Faced with somewhat of a stalemate, given the typical next steps were to start with an AI Pilot, we decided to propose a more simple and familiar solution, and that is to set up a system based solely on ChatGPT.

CharGPT and other LLM's provide the engine room for most AI Agents anyway, and recent developments and tool inclusion have seen them become inceasingly an AI toolkit in their own right.

We proposed that we could devise a system and framework where we could maximise the usefulness of the familair LLM, build policy and process around its use, to the satisfaction of everyone.

The longer term advantage is that when the company decides to scale up, they will already have an advanced grounding in AI, policy and process in place, and many of the CustomGPT's will connect or easily transfer to an AI Agent system.

This proposal was accepted, and here's what we did.

Step 2 – Building a safe framework

We moved the company to a ChatGPT business AI account, created a clear AI policy, and set up an AI working group from across work groups.

Their role:

• Communicate changes to staff, especially the desire of management to enrich jobs, not replace them.

• Consult with Management on the AI Policy development and implementation.

• Approve safe first-round use cases.

• Make sure every AI-assisted job is checked and signed off by a human.

Step 3 – First wave of AI assistants (CustomGPT's)

We built role-specific AI assistants, some of which are:

• Safety Doc Assistant – drafts JSAs and toolbox talks from templates, flags missing controls, and records supervisor sign-off.

• Certification Tracker – scans registers for training/equipment expiries and drafts reminder messages.

• Bid Pack Builder – assembles method statements and risk registers from an approved library for new work proposals.

• Hazard Trend Reporter – summarises incident data for monthly safety reviews.

• Marketing Guru - automatically compliles highlight pictures, video narratives of projects, compiles and sends a fortnighly newsletter to stakeholders and clients.

All work was logged, version-controlled, and only sent out once approved by the nominated human authority.

Step 4 – Results

• Shadow AI use dropped; staff moved to the official system, connected ipads were provided for site supervisors.

• Faster first drafts for safety and bid documents.

• More consistency in language and format.

• Traceability in decisions, approvals and certifications.

• Leaders gained confidence with a clear “Discover → Trial → Scale → Govern” path.

The key takeaway

Treat tools like ChatGPT as company equipment, not personal gadgets:

• Use official accounts.

• Set guardrails.

• Assign responsibilities.

• Keep humans in charge of final decisions.

This company went from divided opinions to a practical, safe rollout - proving baseline tools like ChatGPT, Gemini etc can be affordable, high ROI business tools when managed well.