Skip to content
PROOF // Continuous-Process Industrial Manufacturing Multi-line food & beverage processing plant · IoT-instrumented production floor · continuous-batch operations
← All Proof

CHAPTER 02 · DISCOVER

We Learn Your Operation Before We Propose Anything

One day, on-site. The plant manager framed the stakes up front. Then we did what we always do: talk to people, watch the work, map it, and rank what we found — not just run a meeting.

A named, mapped list of every repeated workflow, by department
A cleaned-up version of the 5 highest-friction workflows
A ranked list of what to fix first, and why
A 30/60/90 plan with a named owner on every line

WHAT WE ACTUALLY DID

We Don't Just Run a Meeting

A roadmap session is where it starts, not where discovery ends. Interviews, documentation, the tools already in use — all of it, before we propose anything.

Interviews

25 sessions / 25 stakeholders

We start with the plant manager, then work down through every shift and department that touches the line — not just the people who signed off on the meeting.

Plant Manager 1× 60min — priorities, budget, what's really at stake
Shift Supervisors (3) 3× 45min — line handoffs and downtime firefighting
Process Engineers (2) 2× 60min — sensor data nobody had time to watch
Maintenance Techs (8) 8× 30min — what actually causes a line to go down
Line Operators (10) 10× 30min — the workarounds nobody wrote down
Quality & Food-Safety Lead 1× 45min — documentation and audit burden

Every session recorded, transcribed, and tagged.

Documentation

58 sources reviewed

Historian exports, SOPs, and the audit records the quality team maintains by hand.

Historian sensor exports vibration, temperature, acoustic — 18 months back
Maintenance logs full CMMS export
SOPs & changeover procedures per line
Quality / food-safety audit records last 2 years
Shift handoff logs paper and shared spreadsheets

Indexed so the agents can reference it directly.

Systems & Tools

7 systems mapped

Everything the floor already touches — no new platform to learn, no rip-and-replace.

Plant historian sensor data system of record
CMMS maintenance
ERP production planning
Quality management system food-safety documentation
Shift-log spreadsheets the real system of record, honestly

All seven connected before we proposed anything.

Floor walkthroughs

We watched the work happen across three shifts and three lines — not just heard about it secondhand.

Process mapping & improvement

Every workflow got mapped end-to-end, then we asked what should change before any of it gets automated.

Adjacent conversations

We talked to the people upstream and downstream of each workflow — not just the person who owns it.

AUTOMATION MAP

Of 165,000 hours a year spent on manual and reactive work, here's how much we can take off the floor's plate

AUTOMATABLE · 112,200 hrs
52,800 hrs HUMAN

Stays human — first-article sign-off, customer relationships, judgment calls on the line.

Of the 112,200 hrs, here's what each workflow contributes

+46K
+31K
+20K
+15K
112K

P1

P2

P3

P4

TOTAL

P1

Predictive Maintenance & Downtime Prevention

46K hrs/yr · $2.6M

P2

Sensor-Data Monitoring & Briefing

31K hrs/yr · $1.7M

P3

Batch-Quality Deviation Tracking

20K hrs/yr · $1.1M

P4

Energy & Utility Cost Monitoring

15K hrs/yr · $0.8M

What we picked to build first

Plant-floor sensor briefing & remediation

Highest-leverage, lowest-risk: the sensor data already existed in the historian, the pain was universal across every line, and a visible catch in the first month would change every later conversation more than a bigger, slower project would.

Committed on the wall. Working session scheduled within 10 days.

16 weeks to the first system live · then ongoing

Discover

Working session, on-site

1w

Sensor briefing & remediation

Wave 1

7w

Predictive maintenance

Wave 1

6w

Batch-quality & energy optimization

Wave 2

6w
Week 1 Week 16 · then ongoing →

THE METHOD

Same Approach, Every Engagement

Most AI roadmaps fail the same way — a list of interesting projects with no way to choose which one matters. Ours turns what we actually find on the floor into a ranked, ownable plan.

01

Discover

What do you repeat? What frustrates you? What runs on a schedule?

→ A named list of the real, repeated work — across every department.

02

Map & Improve

How does this actually happen, workarounds included — and what should change before we automate it?

→ An end-to-end picture of each workflow, cleaned up first.

03

Rank & Commit

What ships first? Who owns it? How do we know it worked?

→ A ranked plan with a named owner and a metric on every line.

CHAPTER 02 OF 5

UP NEXT · BUILD

See the agents we actually built and shipped — including the exact steps one of them runs in production.

Continue to Build