The cost of reacting late
By the time an operator notices a drift and adjusts, product is already off-spec. Manual correction every 15 minutes, on every grade/SKU, on every shift, isn't a job description anyone signed up for.
Autonomous optimization, ready when you are.
ProcessMiner begins every deployment with real-time quality prediction, then specific recommendations your team can verify. From there, the open-loop concept and supervised closed-loop let operators keep full control, and only when a customer is ready do we move their lines to autonomous, real-time setpoint optimization. Closed-loop control isn't just possible, it's proven.
The real tension
Constant manual correction isn’t sustainable.
Blind trust isn’t realistic.
Operators have competing priorities. They can't stop to re-optimize a setpoint every 15 minutes on top of everything else on the floor, which is exactly why so many process quality deviations get caught late instead of prevented. But the answer isn't handing control to an AI platform overnight. Process engineers who've spent years learning a line's quirks have good reason to be cautious about a system making line changes without being in the loop. For that reason, we start with the open-loop concept and supervised closed-loop, where the operator keeps full control.
The cost of handing over control blind
An AI platform making live changes to a process you can't see or override erodes trust fast, especially on equipment where a wrong setpoint has real consequences.
The middle path that actually works
Start with recommendations you can see and verify. Move to closed-loop control only on the variables, and for the hours, you're ready to trust it. Scale from there. This isn't about replacing anyone. We're adding a tool to an already-expert operator that helps them do their job better — an upskilling tool that comes alongside your team.
Advisory mode vs. autonomous mode
The Best of Both Worlds
Every deployment runs on the same predictions and centerlining guidance. The difference is who executes the correction: your operator, following a specific recommendation, or the platform, writing the setpoint directly, only where you've authorized it.
Advisory mode
You act on the recommendation
The platform tells you what's drifting and by how much; your operator makes the change.
- Trend flagged. The platform surfaces the deviation and the specific setpoint change to make.
- Human decides. Your operator reviews and applies the change, or overrides it, on their timeline.
- Result depends on response time. Value is real, but capped by how fast a person can act between other duties.
Autonomous mode
The platform acts, within limits you set
The same prediction triggers a direct setpoint write, on variables and hours you've authorized.
- Trend predicted. The same 30-second model forecasts the deviation before it reaches spec.
- Correction written automatically. Only on variables you've cleared for closed-loop control.
- Operator retains the switch. Autonomy can be toggled off instantly, any time, for any reason.
Proof point One deployment, chemistry usage
~1,600 lbs
Chemistry saved in one week
At only 32% autonomous usage
32%
Of the week in autonomous mode
The rest stayed manual
>20%
Quality variability reduction
Pulp & paper deployments
$2.4M/yr
Value on a single paper machine
50% attributed to platform by customer
Figures reflect specific ProcessMiner deployments. Actual results vary by facility, process, and data quality; past performance is not a guarantee of future results. The $2.4M/year figure is per machine, not an aggregate claim.
The adoption path
Four stages. You control every transition.
No deployment starts in closed-loop. Trust builds in stages, each one gated by your team's confidence in the recommendations, not by a sales timeline.
Open-loop recommendations
The platform recommends specific setpoint adjustments. Your operators evaluate and apply them manually, building a track record you can verify against your own results.
Limited closed-loop
Autonomous control runs 1–2 hours per day, typically during stable, well-understood production windows, with operators watching and free to intervene.
Expanded closed-loop
As confidence grows, autonomous windows extend to 8+ hours a day, covering more of a shift while still leaving room for manual override during transitions.
Full autonomy
Closed-loop control runs continuously on the machine, with operators retaining the ability to disable it instantly at any point, for any reason.
What you control, always
Autonomy has an off switch. Literally.
The platform doesn't decide when it's ready. You do. That decision is enforced at the variable level and at the moment level, not as a policy promise but as a setting your team operates directly.
Variable-level control
- You choose which process variables are eligible for autonomous adjustment, for example chemistry dosing, and which stay strictly manual, for example basis weight targets during a customer-specific run.
- New variables are added to closed-loop scope only when your team decides the track record supports it.
- Every autonomous write is logged against the recommendation that triggered it, so any change is traceable back to its cause.
Moment-level control
- Operators toggle autonomy off during a grade transition, for example moving from Grade A to A1 to A2 to B, and re-enable it once the process stabilizes on the new grade.
- Autonomy can be switched off instantly from the floor, for a startup, a maintenance window, an unfamiliar disturbance, or no stated reason at all.
- Closed-loop writes connect through IoT-agnostic Modbus TCP to your DCS, historian scripts, or API write-back, whichever fits your existing control architecture.
Autonomous control acts on the same golden-run guidance that powers Intelligent Centerlining, and decisions are triggered by the same 30-second predictions your team already sees in Real-Time Prediction.
Put us to the test
Prove it on your data.
Start with advisory recommendations on your process, no autonomy required. We'll show you the trend accuracy and the specific setpoint guidance before any conversation about closed-loop control even starts.
Autonomy is optional and gradual. You decide if and when to move past advisory mode.
Questions, answered
Frequently asked questions
Can ProcessMiner actually control our equipment autonomously?
Yes. Deployments typically start with advisory dashboards and centerlining, then can graduate to closed-loop autonomous control when desired. You retain complete control over which variables are automated versus which stay manual.
How long does it take to reach full autonomy?
This varies by facility. One client’s journey to running roughly 20 machines in autonomous control took about two years, built one proof point at a time. There is no fixed enterprise-wide timeline, and mill-level decisions are made at the mill level.
What happens during grade transitions?
Operators can toggle autonomy off during a grade transition (for example Grade A to A1 to A2 to B) and re-enable it once the process stabilizes on the new grade.
How does the system connect to our control systems?
Closed-loop integration is flexible: IoT-agnostic Modbus TCP to your DCS, historian scripts, or API write-back. Open-loop recommendations are the default starting point for every deployment.
Do we need to go straight to full autonomy to see value?
No. One deployment saved roughly 1,600 lbs of chemistry in a single week using autonomous mode only 32% of the time. Partial adoption still delivers measurable savings well before full autonomy.
Get in touch
Have a question?
Whether it's our platform, your process, or a free in-depth process data analysis and pilot you'd like to see in action, let us know.