Proof, on their scorecards

We don't pitch results. We point at them.

Real operations, real scorecards. Measured before, measured after, in their numbers and not ours. Here's what that looks like.

2 weeks by hand
2 hours
Lease data into Yardi
1+ hour every night
Hands-off
Nightly crew routing
~1 day a month by hand
0 hours
Program data, monthly
2 hours partners only
5 minutes
Client report build
900 NORTH MICHIGAN · Real estate & property mgmt

108 leases into Yardi. Zero errors against the source.

The terms that run the business, rent, escalations, renewal dates, options, were locked inside 108 lease documents and aging legacy systems, and every one had to land in Yardi correctly. By hand, each round of collecting and transforming this data would have taken weeks. One mistyped escalation is a financial miss that surfaces months later.

We built the workflow that reads the leases, pulls the structured terms, and loads them into Yardi, with every field checked back against the source document. AI does the reading. The check is what makes the output safe to trust.

What changed for them. The data stopped being a project. New leases run through the same workflow in hours, and the team spends its time on the exceptions instead of the typing. When the workflow gets something wrong, it surfaces in the check, not in a rent roll six months later.

"We were completely stalled, trying to get data out of legacy systems. Now we do in hours what used to take weeks."
900 North Michigan · Real estate & property mgmt
On their scorecard · Data extraction time
2 weeks by hand, per round
2 hours
Leases processed
108
Accuracy vs source
100%
JUNK RELIEF · Home services · Chicago

Twenty years of routing by hand. Now his own automation does it at 6 p.m. every day.

JUNK Relief is a Chicago junk-removal company whose owner had routed the next day's crews by hand for twenty years, an hour or more a night, because he believed nobody could do it better. Its website ran on WordPress on an outside agency's stack he did not control.

Over roughly three weeks, Cylentra put him on a Claude Team plan, trained him, and got him Service Fusion access, then he built the routing automation himself over one night of working with Claude. Cylentra rebuilt his website like-for-like onto Cylentra hosting on Azure and cut it over live, keeping the old site as a one-step rollback.

He was not replaced by the tool. A man who had done a job by hand for twenty years reached, in one night, something he had assumed would take a month.

"It routed it more intelligently than I would ... I've done this for 20 years. I spent an hour a night if not more routing these things because nobody can do it better than me. And in one hour of me training this thing it will now do it for me at 6 p.m. every night. Hard stop. I'll never have to do it again."
Joe Weidman, Owner, JUNK Relief
On his scorecard · Nightly crew routing
1+ hour every night, for twenty years
Hands-off
Runs itself at 6 p.m. daily. He says it routes better than he did by hand.
Website platform
agency WordPress → Cylentra hosting on Azure
Built by
the owner himself, in one night
CANINE THERAPY CORPS · Nonprofit · Chicago

A dozen spreadsheets, totaled by hand. Now one source of truth.

Canine Therapy Corps is a Chicago nonprofit that brings therapy dogs into hospitals and rehabilitation programs. Its small team ran on scattered data: therapy hours lived in about a dozen spreadsheets, finances lived in QuickBooks, and donor records lived in Salesforce. None of it talked to the rest, so every number had to be assembled by hand. The time went into collecting the data, which left none for asking what it meant.

Over four weeks, Cylentra trained the team to build their own Claude skills against Canine Therapy Corps' real systems, rather than handing over finished software. The Executive Director built a QuickBooks actuals-vs-budget-vs-prior-year board report and a one-click donor-metrics view. A staff member built Salesforce skills for data imports and monthly reconciliation. A third built a scheduled monthly transfer that moves attendance data from Google Sheets into Salesforce automatically, about 2,000 records a month.

What changed. Totaling program-partner hours used to take the Executive Director the better part of a day each month, across a dozen spreadsheets. Now the attendance data lands in Salesforce on a schedule, about 2,000 records a month, and the total is a question to Claude. Staff kept logging work in the spreadsheets they already knew. The Executive Director expected pushback from the team and saw very little of it.

"Now all I have to do is ask Claude. Across everything, especially for the way that I work, it is going to not only save me a lot of time, but allow me to do a lot of things I was never able to do before."
Sheila Gidley, Executive Director, Canine Therapy Corps
On their scorecard · Program-partner hours
~1 day a month, totaling 12 spreadsheets by hand
0 hours
Hand tally gone. ~2,000 records a month land in Salesforce on a schedule.
Board reporting
two docs, a walkthrough → one report
Answering a data question
hours of hand assembly → ask Claude
STRIDE · Financial advisory

The method lived in the partners' heads. Now it lives in the system.

STRIDE's partners had a report process perfected over years. It lived in their heads, and only they could run it. Every client report pulled a partner off partner work, and the firm could only produce as many reports as the partners had hours for. That's the ceiling: not a market problem, a key-man problem.

We captured the methodology into a knowledge base and a set of templates, then built the workflow on Claude that runs them. An advisor supplies the client data, Claude drafts the analysis against the partners' own method, and the partners review instead of build. Partner-level analysis isn't locked to the partners anymore.

What changed for them. The partners got their hours back for partner work. The workflow has been in production with STRIDE's advisors since spring 2026. A new advisor produces partner-quality analysis in their first month, because the method is written down instead of remembered. Nobody re-explains it. Fixed once, and it stuck.

"We wanted to capture our methodology so the whole team could produce partner-level analysis, not just the partners."
STRIDE · Financial advisory
On their scorecard · Report build time
2 hours by hand, partners only
5 minutes
Any advisor, partner-level, in the system.
Who can produce them
partners only → any advisor
Output, report after report
inconsistent → partner-level
HISTORICAL ARCHIVE · Research archive · Independent project

~8,700 pencil marks found. ~13,000 passages classified. Now headed into print.

A privately held archive holds a several-thousand-page reference work, hand-annotated in pencil by one domain expert over years, flagging passages of significance across tens of thousands of entries. A prior manual catalog captured most of it, but nobody could say how much it missed. The gaps were never measured, so "complete" was a guess.

A computer-vision pass tuned for faint pencil marks processed every page, then was audited against the prior catalog to measure real recall instead of assuming it. Classification ran across dozens of parallel AI agents down to individual senses and sub-entries, one of twenty categories each, every call backed by a quoted phrase and every batch validated for completeness. All of it flowed into a browser-based review tool with one-click override. An independent project we built to prove the method under real conditions, not a paid engagement.

What changed. The expert's judgment stayed load-bearing, setting the rules, deciding the edge cases, and approving every override. What changed was scale, full classification depth across the entire archive instead of a manual catalog that could never prove its own coverage. There is a built-in way to prove how complete the result is, and the classification is now slated for publication as a supplemental section of a scholarly reference book.

"You don't know what you can do with AI until you do it."
Project owner · Historical Archive · Independent project
On the scorecard · Annotations catalogued
Unknown coverage, one expert by hand
~8,700 found
Every page audited for recall. ~13,000 passages classified, headed for a scholarly reference book.
In their words

The people who run these operations, on what changed.

Every quote below is from the operator whose scorecard moved. Read the full story behind any of them.

"Now we do in hours what used to take weeks."
900 North Michigan900 North Michigan · Real estate & property mgmt
"And in one hour of me training this thing it will now do it for me at 6 p.m. every night."
JUNK ReliefJoe Weidman, Owner, JUNK Relief
"Now all I have to do is ask Claude."
Canine Therapy CorpsSheila Gidley, Executive Director, Canine Therapy Corps
"We wanted to capture our methodology so the whole team could produce partner-level analysis, not just the partners."
STRIDESTRIDE · Financial advisory
"You don't know what you can do with AI until you do it."
Project owner · Historical Archive · Independent project

Your operation is next.

Book a call. We'll find the first workflow the engine should be running in your business, put it on a scorecard, and prove it in 60 days.

Book a call

Client results shared with permission.