Thirty days at an engineering firm. Including the two zeros.
An engineering firm ran a 30-day proof of concept with me: one contracted workflow, technical reporting, five seats, and a written report at day 30. This page is that report with the firm's name and its people's names taken out, and nothing else taken out.
What the letter asked for, and what landed.
One workflow, built and proven inside the reporting team
Delivered three, each validated against live work. Profile sheets for one bridge job (4 sheets, 39 borings, 13 lab-classified samples) and for a second job nobody at the firm had started. Drilled shaft tables run against the engineer's own answer keys on three borings, one of them blind. The front end of a technical report, two of its six sections, tested against a report the firm had already issued.
Five employees onboarded, with training
Six. Kickoff in the first week, individual install sessions the same week, a personalised audio brief per role.
Check-ins and measurement against week-one baselines
Baselines were captured from four of the six before their installs. The other two touched the tool first, so their day-30 answers are reported standalone rather than as a before-and-after. Stated plainly, because the next phase fixes it with timed measurement instead of survey estimates.
A day-30 report: results, survey, a go or no-go, a next-phase scope
Delivered at day 30. Recommendation: refine. Not continue. The reasons are below.
Outside the letter: drainage
A drainage tool that saves 20+ hours of manual work per job, by the engineer's own numbers: model tabulation, inlet tables, and permit-form entry that were being retyped by hand from outputs the model already held. Reviewed by the firm's drainage engineer against a completed submittal.
The finding that mattered most was accuracy, not speed.
For a firm whose output is sealed and submitted, being right matters at least as much as being fast. None of this was required by the letter.
An error found in completed work
On one boring, the reconciliation check flagged a termination depth in a finished appendix that did not match the boring record: 75.6 feet against 83.6. The engineer confirmed the record was right. The firm's own review had caught it independently. The automated check and the firm's QC agreed without talking to each other.
The same checking surfaced five defects in the logging software's export itself. The engineer took the ones that mattered to the vendor, and the vendor came back asking for more detail.
A correlation calculator, rebuilt from the standard
The drilled-shaft calculator was rebuilt from the FHWA source standard rather than copied from the firm's workbook. It matches the engineer's own calculator to nine decimals on every cell, on two workbooks two weeks apart.
Report sections against an issued report
Boring counts, boring IDs, and the standard technical language came back as exact matches. One depth figure differed, because the tool counted coring depth and the firm does not. The engineer settled the convention in one line, and it is built in.
An export check on a forty-boring job
40 borings, 0 disagreements against the firm's workbooks on every row the check could compare, with the count of rows it actually compared printed beside the result.
The day-30 survey, published whole.
Six role-matched forms went out at day 30, each mirroring that person's own baseline questions plus five closing questions. Five came back within a day. Hours are each person's own estimate for the month, not timed. Nothing on this page adds them up.
| Role | Hours saved, last month | Correction needed | Keep using it | Recommend, 0 to 10 |
|---|---|---|---|---|
| Operations lead | 20 | Minor edits | Yes, with changes | 5 |
| Data analyst | 18 | Minor edits | Yes | Answered off the scale; read as high, not counted |
| Engineer | 0 | "I did not use it" | Yes | 5 |
| Engineer | 0 | Significant rework | Not sure | 3 |
| Finance lead | 60 | Minor edits | Not captured | Not captured |
| Practice lead | Not received | |||
The three who ran the system on their own work, unsupervised, reported hours back with minor edits. The two engineers, who received finished output built on my machine, reported zero, and asked in different words for the same thing: put the tool in my hands. That is the design of the pilot showing up in its own numbers. Every geotech job ran on my side to protect accuracy on sealed work, which proved the mechanism well and measured the engineers' hours poorly. The finance lead's figure is the largest in the pilot and stands as his self-report; at kickoff he named the evidence that would convince him the program paid for itself, and the day-30 report was scored against that test.
What the pilot did not produce.
An engineer-measured time saving. The one measurement that exists: a project designer new to the firm finished one generated profile sheet in 6 to 8 hours, starting from the delivered drawing file, and named four adjustments, every one a house drafting convention the tool had not been given. There is no fair baseline to set that number against, so the report claims no saving and rules none out.
On the drilled-shaft tables, a boring the system had never seen, volunteered blind by the engineer, came back 41 of 97 cells exact after two of his rules were written down. Every miss is attributed to a named cause, most of them judgment calls he tunes by hand. That result is why the recommendation was refine, not continue.
The firm's standard was not written down anywhere.
Seven drafting and reporting conventions surfaced in thirty days: which N-value prints beside a boring stick (three answers inside the firm), how a report states its depth range, what label sits below a terminated boring, which of two blow-count fields governs, what a zero groundwater reading means, hatch patterns and line weights, and text style and stick width. Each was held by a different person. None was in any file the firm had.
Each is now a written, tested rule with a name and a date on it, carried by the tools. The firm had never had to write them down, because the people who hold them do the drawing. That standard did not exist on the day the engagement letter was signed. It is as much the product of the thirty days as the tools are.
What people did, with the clock running.
Twenty-two minutes
An engineer forwarded the first generated sheets to a colleague 22 minutes after receiving them. He has since sent his own fourteen-item data specification to the logging-software vendor, unprompted.
The day after the zero
The engineer who answered "I did not use it" installed the drilled-shaft tool on his own machine the next day and ran it on a live boring. He dictated a cell-by-cell review that found two real bugs and six of his own rules the tool did not yet know. All of it was fixed the same evening, and the rebuilt tool reproduced his six skin-friction values from his dictated inputs. First engineer-operated production run of the engagement.
Six days after install
The chief operating officer filed a bug report.
Outside the six seats
A designer who was never in the pilot is finishing generated sheets and timing the slow steps to report back.
Refine, not continue.
The decision rule was fixed in advance, before any verdict existed. Continue, if the engineer's verdict is that the output saved real time. Refine, if a rebuild was required, with the stated reason becoming the next round's specification. Stop, if the verdict is negative on mechanism rather than calibration. A rebuild was required and the designer stated the reasons, which is the refine branch by its own wording.
The recommendation: a 90-day scale phase gated at day 30 on the same job, the same engineer, and the same measurement, built so the engineers produce the number themselves. Running every job on my machine was the right call for proving the mechanism and the wrong call for building the habit. Their hands, their job, their machine, timed.
The 30-Day Proof runs the same way at your firm.
One workflow. Up to five people. A measured before and after, and a written report you keep whether or not we continue, including the parts that did not work.
Book a 15-minute fit call →Not sure where you stand? The Readiness Survey takes ten minutes.