Real workflows, real baselines, measured results.
Client names withheld by agreement; numbers are actual.
A solo-run racking pipeline, now running on AI workflows the company owns.
A solo-managed sales and delivery pipeline. Quoting, tender review, proposals, and collections all handled manually, by one person.
Each tender read line-by-line. Estimates built from scratch. Proposals assembled manually. Overdue invoices chased ad hoc.
A production AI workflow system: tender dissection with scope-gap flags, an estimating engine with company margin rules, branded proposal and quote generation, staged collections sequences, and LinkedIn and client-comms drafting.
Timed against real jobs, not samples. The same operator built quotes the old way for four weeks while the baseline was recorded, then the same job types were rebuilt through the new workflow. Reported figures are the median, not the best run.
Pricing authority, margin decisions, and the final read on every quote stayed with the operator. No supplier relationships were altered. No engineering was performed or replaced — stamped drawings still come from the engineer of record.
How the numbers on this page were produced
A time saving is only worth reporting if the before number was measured rather than remembered. Every engagement therefore opens with a baseline period: the existing workflow is timed as it actually runs, on live jobs, including the interruptions and the waiting.
These are workflow results, not sales results. Faster quoting does not by itself win more work, and nothing on this page should be read as a revenue promise. Client names are withheld by agreement; the figures are actual and the clients have seen this page.
Ten hours of admin a week, per rep, down to one.
A large contractor wanting AI inside the systems it already ran, not alongside them. The sales team was absorbing the administrative load itself, and roughly a quarter of each rep’s week was going to it.
No triage layer sat in front of the work. Every administrative task arrived at the same priority and was worked by hand, in the order somebody noticed it, by whichever rep it landed on.
A triage layer and a set of agentic workflows, deployed inside the client’s own environment to their IT and data-handling requirements. Incoming admin is classified, routed, and drafted by the agents. The reps now review output instead of producing it.
Partly logged and partly reconstructed from the team’s own records — a weaker basis than Case 01, and worth saying so. The figure below is the sales team’s own reported weekly admin load before and after, not a timed study.
Engineering sign-off and final review stayed in-house, unchanged. Nothing an agent produces reaches a client until the responsible person has read it and approved it.
Two figures, not three. The third would be padding.
“This has been a gamechanger for our sales team.”
Two further design-partner engagements are in their baseline period. Results are published once the after figures are measured, not before.
Every engagement starts with a baseline.
Before anything gets built, the current cost in hours is measured. That baseline is what the results above are compared against, and it is the first thing a Workflow Audit produces.
AI workflow consulting for material handling businesses — racking dealers, installers, and distributors across Ontario. Twenty years on the floor, applied to the paperwork.