Key Takeaways
- HMA is the operating model: HMA means humans, machines and agents working together, and it decides what each one does. It runs on the systems a food and beverage plant already has, and any module can be replaced later.
- Forward deployed experts find the plant's know-how, and NaturalAI™ keeps it: InspireXT domain experts work on your plant floor beside supervisors, planners and buyers, from discovery through go-live, because the workarounds that break AI agents are only visible there. NaturalAI™ records every rule, workaround, agent limit and decision against the process it belongs to, and anyone can query it in plain language.
- Imperative one, expose silent compensation: silent compensation is the manual work people do every shift to make the ERP match the plant. Every food plant runs two supply chains, the one in the ERP and the one buyers and supervisors keep alive with undocumented workarounds. Find and fix those workarounds before any AI agent goes live.
- Imperative two, take the screen out of execution: on a connected shopfloor, machine signals reach the ERP as work order updates, quality holds and maintenance requests, without a supervisor retyping them. Machines report, agents execute, and people govern the exceptions.
- Imperative three, make agents act only on live data: under the Fresh Start rule, an agent reads current values from the system of record before every action, and never works from a cached snapshot.
What is a connected shopfloor in food and beverage manufacturing?
A connected shopfloor is a food and beverage plant where machine data flows straight into the systems that run production, maintenance and quality, so a reading on the line becomes a transaction in the ERP without a person retyping it.
Most food plants already collect plenty of machine data. It sits in PLCs, SCADA screens and historians, visible to the engineer on shift and invisible to the planner, the buyer and the quality manager. A connected shopfloor closes that gap in three steps.
Step 1: Connect the shopfloor.
Machines, PLCs, SCADA systems, sensors and historians are linked through an industrial edge platform. It collects their readings, normalises the formats and adds business context: which line, which batch, which product. A raw temperature value of 4.2 becomes holding tank 3, batch 4471, tomato sauce, within limit.
Step 2: Execute the ERP.
The contextualised signals feed manufacturing execution, maintenance and quality workflows. A completed fill count becomes a production completion. A vibration threshold breach becomes a maintenance work order. A failed inline check becomes a quality hold with the batch genealogy already attached.
Step 3: Scale Connected Intelligence.
Once machine data and enterprise data share one governed foundation, the plant can build dashboards, predictive models, AI agents and apps on top of it. This is where the line connects to demand, inventory and planning. NaturalAI™ adds the process context that tells each reading which process it belongs to and who owns it.
A plant that skips step 1 builds AI on data nobody trusts. A plant that skips step 2 ends up with excellent dashboards and the same manual transactions.
How does the HMA model divide work between humans, machines and agents?
The HMA model gives each party the work it does best: machines report physical reality, agents carry out the transactions that follow, and people decide what agents may do.
The machine: physical execution and edge data. Sensors, PLCs and edge gateways stream readings every second or faster. They are the plant’s physical record of what is happening on the line.
The agent: digital orchestration and execution. Software agents read those signals and run multi-step ERP transactions through APIs, such as an inventory transfer, a revised routing or an adjusted bill of materials. Every action goes into an audit log.
The human: governance and exception management. The supervisor stops keying transactions and starts governing. People set the monetary and volume limits, approve changes above a threshold, and take the cases that need judgement, a sensory check or a call to a customer.
The rules that connect the three have to live somewhere. In an InspireXT deployment they live in NaturalAI™, tied to the process each one belongs to: which line, which product, which approval limit and which named person signs off. When an agent halts, the halt is recorded against that process, so the team can see which workaround caused it and who owns the fix.
HMA does not depend on any one vendor. Underneath it sits a headless architecture: the ERP keeps all of its business logic, but its screens stop being the only way in. For a food and beverage producer the practical test is one question. Can every action a supervisor takes on a screen also be taken through an API?
The platforms underneath can change. What has to stay constant is the knowledge of how the plant actually runs. That is why InspireXT works forward deployed, on NaturalAI™: engineers on the floor find that knowledge, and NaturalAI™ keeps it.
| What changes | Before HMA | After HMA |
|---|---|---|
| How transactions happen | People key them into ERP and MES screens | Agents call ERP and MES APIs |
| Time from event to record | Minutes to a full shift | Seconds |
| Workarounds | Made by hand and never recorded | Surfaced as exceptions and logged against the process |
| Role of AI | Copilot summarises and suggests | Agent executes within set limits |
| Software model | One suite, upgraded as a block | Composable modules, replaced one at a time |
How does InspireXT deliver a connected shopfloor?
InspireXT delivers a connected shopfloor with forward deployed domain experts who work on the client’s plant floor, using NaturalAI™ as the base for everything they build.
Forward deployed: domain experts on your plant floor
Forward deployed means our domain experts work inside your operation, beside your supervisors, planners and buyers, and build with them. They do not design from a distance and hand over a solution.
This matters for one reason. The workarounds that break AI agents are only visible on the floor. An engineer standing at the filler on line 2 sees the extra film being issued. A remote team reading the ERP sees a standard fill weight and nothing wrong. The forward deployed team records each workaround as it finds it, fixes the master data behind it, and then builds the agent with the supervisor who used to carry it.
The same people stay through go-live. When the first agents start halting, the engineers who found the original workarounds are on site to read the halts and fix the cause.
NaturalAI™: the base the domain experts bring
NaturalAI™ is InspireXT’s AI-native platform for process-led transformation. It brings ten years of InspireXT process IP, a repeatable forward deployment method, automation and AI agents into one environment. On a shopfloor it does three jobs:
- It starts the team from the Connected Process Model (CPM): the domain experts map the plant’s current state against it, finding process gaps, decision gaps and data gaps, so no time goes on drawing processes from scratch.
- It holds the rules agents follow: every limit, approval threshold and named owner is tied to the process it governs, and Conversational Process Intelligence lets supervisors and managers query them in plain language from week one.
- It gives machine data process context: a data platform shows that fill weight drifted. NaturalAI™ shows which process that drift belongs to and who owns the fix. When an agent acts, NaturalAI™ records which rule it followed and who approved it, so any decision can be traced back in plain language.
The work follows the four phases of the Connected Delivery Method: Discover, Blueprint, Transform and Manage. Discover, the phase where silent compensation is found, typically takes four to six weeks.
Why must AI agents act only on live data?
AI agents must read current values from the system of record before every transaction, because an agent working from a cached snapshot commits stock and capacity that no longer exist.
We call this the Fresh Start rule. On a busy chilled line, a ten-minute-old inventory snapshot can allocate sauce that line 2 has already filled and packed, and the failed order then ripples through picking, dispatch and invoicing.
Fresh Start has a second effect leaders should prepare for. When the data and the plant disagree, a well-built agent stops and asks for a person. In its first weeks it will stop often. Each stop marks a place where someone has been compensating silently. The forward deployed team logs every halt in NaturalAI™ against the process it came from, groups the halts by cause, and fixes the master data behind each group. The result is the most accurate map of the plant’s data the business has ever had.
Scenario: a temperature excursion on a food production line
This is an illustrative scenario built from common patterns in food production.
About the plant: Adavia Foods is a mid-sized food manufacturer making chilled sauces and ready meals on four lines, running a cloud ERP, with its machine data locked in SCADA screens and a historian.
The challenge: at 02:10, holding tank 3, with a batch of cooked tomato sauce waiting to be filled, begins to warm after a chiller fault. The sauce must stay below 5°C. The sensor registers it at once. The supervisor notices at 02:40, walks to a terminal and works through the work order, batch record and inventory screens to decide whether the sauce can be diverted to the hot-fill line, where it will be cooked again. The decision is keyed in at 03:30, and nobody records why it was made.
What changes: on a connected line, the drift is picked up at 02:10 and published, tagged with the tank and batch. NaturalAI™ already holds the rule for this process, agreed with the quality manager during Discover: chilled sauce above 5°C goes on hold, can only be diverted to hot-fill, and transfers of this size need the night supervisor’s approval. A business rule raises a production exception and places the batch on hold. An agent checks open orders, finds a hot-fill run at 04:00 that can take the sauce, and prepares the inventory transfer, work order update and bill of materials change. At 02:14 the supervisor approves it from a single message.
Value delivered: response time falls from eighty minutes to four. NaturalAI™ records the decision against the process, with the rule the agent followed and the person who approved it, so at 08:00 the quality manager can ask why batch 4471 moved to hot-fill and get the answer in plain words. Every move is also logged with its lot code, quantity, location and time. Fresh tomatoes are on the FDA’s Food Traceability List, and because cooking is a kill step, the FDA Food Traceability Rule asks the plant for records of each tomato lot received and of the cooking step. The agent’s log links both to batch 4471. The FDA will not enforce the rule before July 20, 2028, and a plant whose agents log each move as it happens already holds the evidence. For UK and EU producers, General Food Law (Regulation (EC) No 178/2002, Article 18, retained in UK law) already requires one step back, one step forward traceability, and the same log meets it.
The three operational imperatives for food and beverage manufacturing leaders
Food and beverage manufacturing leaders remove the screen bottleneck by exposing silent compensation first, moving execution to APIs second, and enforcing live data throughout.
- Expose silent compensation. Run a master data diagnostic on the floor before any agent goes live. During Discover, put engineers beside buyers, planners and shift supervisors for two weeks and record every manual override. Compare actual lead times, yields and fill weights with the values in the ERP. Record each one in NaturalAI™ against the process it affects, so it becomes a fix the whole team can see. Every gap closed now is a halt avoided later.
- Take the screen out of execution. Write API access to every ERP and MES transaction into new software contracts. Start with one high-friction flow, such as rerouting product after a temperature excursion, and give the agent firm monetary and volume limits. Move experienced supervisors into governing roles, because the people who know the workarounds are the best people to set an agent’s limits. Write those limits into NaturalAI™, where each one sits with the process it governs and the person who approved it.
- Make agents act only on live data. Put Fresh Start into IT policy. No agent, copilot or scheduled workflow transacts on cached or batch data, and any agent that meets contradictory readings halts and escalates to a named person. Each halt is logged in NaturalAI™ with the process it came from, so the pattern of halts becomes the plant’s improvement list.
Most plants find their first agent stops more often than it acts. Each stop is a workaround someone has carried for years, written down at last, and once enough of them are fixed the ERP finally describes the plant you actually run.
Frequently Asked Questions
What is the HMA model in food and beverage manufacturing?
HMA stands for humans, machines and agents. It is an operating model in which machines report what happens on the line, AI agents carry out the ERP transactions that follow, and people set the limits and handle exceptions. It runs on the edge, ERP and data platforms a plant already has, and NaturalAI™ keeps the process knowledge separate from them, so any module can be replaced later.
What is NaturalAI™ and what does it do on a shopfloor?
NaturalAI™ is InspireXT’s AI-native platform for process-led transformation, built on ten years of InspireXT process IP. On a shopfloor it holds the process model of the plant, the rules and limits each agent follows, and a record of every workaround, halt and decision. Supervisors and managers can ask it questions in plain language, such as why a batch moved or which rule an agent used.
What is the difference between an AI copilot and an AI agent in manufacturing?
A copilot reads data and suggests an action, and a person still opens the screen to carry it out. An agent carries out the transaction itself through the ERP’s APIs, within limits a person has set.
Do we have to replace our ERP to use AI agents?
No. Agents reach the ERP through APIs, and its business logic stays where it is. An integration layer can expose an existing ERP’s transactions this way, so most food and beverage producers start with the ERP they have and replace individual modules later.
How does a connected shopfloor help with FSMA 204 and UK traceability rules?
Each agent transaction records the lot code, quantity, location and timestamp as it happens. That builds the traceability records the FDA’s rule requires for foods on its Food Traceability List, and the one step back, one step forward records UK and EU General Food Law requires, continuously, so nobody has to assemble them after a request arrives.
Book a Connected Shopfloor Readiness Audit.
Our engineers will find the silent compensation in your master data, record it in NaturalAI™ against the process it affects, and show you which flow your first agent should take, before you deploy it.