AI Doesn't Need More Process Maps. It Needs Process Atoms.

Welcome back to Work in Process, a newsletter for senior process, transformation, and enterprise architecture professionals who are accountable for making AI work inside large organisations.

This week: SAP introduced Process Atoms to support AI agents; Lucid set out a practical AI readiness checklist across four organisational dimensions; Siemens and PepsiCo showed what happens when the knowledge layer comes before the AI; and Samsung is shipping Ballie, a home AI companion robot powered by Google Gemini.

BPM has spent thirty years building tools to describe processes. Yet it has lacked a practical way to bring together the context, rules, and live data that AI needs to act.

Process models show the flow but not the meaning. Event logs show what happened but not why. Business rules handle decisions but stay disconnected from the actual workflows. For decades, experienced humans filled these gaps using their own context and judgement.

AI agents cannot do that. They need to know which rule applies, when, for which customer, and what happens if a constraint is breached. None of this lives in a standard BPMN diagram.

SAP calls the unit that holds this data a Process Atom. It is a clear, context-bound rule that tells an AI agent exactly what must happen and when. It does not replace BPMN or event logs; it connects them so AI can act.

A business rule simply checks a condition and gives a decision. An atom governs behaviour over time, fits a specific scope, and can be tracked against real data. For example, "Orders over £10,000 need manager approval" is a rule. "Urgent, high-risk orders over £10,000 need a senior analyst review within 2 hours" is an atom. It adds time, roles, and context that can be measured against what really happened in the system.

This knowledge already exists in your organisation. It lives in your diagrams, policy documents, SOPs, and the minds of your staff. However, it has never been unified, structured, and machine-readable in one place.

Your process team is one of the few groups that already has access to all of this information. The opportunity now is to structure it in a way that AI can understand and use.

Read more: SAP — Process Atoms White Paper ; SAP Community — Process Atoms: Towards AI-Powered Business Process Management

Lucid built an AI readiness checklist covering four organisational dimensions.

Written by Lucid Solutions Consultant Patrick Saul with agentic AI specifically in mind, the checklist assesses organisational readiness across processes, data, systems, and culture. It positions AI readiness as something to evaluate before deployment rather than after, and is aimed at operations and transformation leaders who need a structured way to identify gaps and build a case for investment.

More tool than theory: it is designed to generate output, not just perspective. The checklist is available to download free from Lucid's website at the link below.

If your team is building a case for AI investment this quarter, this gives you a structured way to identify where the gaps are and name them.

Read more: Lucid — AI Readiness Assessment

PepsiCo built the knowledge layer before the AI. Throughput rose 20%.

Siemens built what they call a "comprehensive knowledge ecosystem", a platform that fuses manuals, maintenance documents, machine controllers, and ERP data into a single structured layer, before introducing AI.

PepsiCo worked with Siemens and NVIDIA to create high-fidelity 3D digital twins of its US manufacturing and warehouse facilities, capturing every machine, conveyor, pallet route, and operator path, before AI agents were introduced. Those agents then simulate and refine system changes against the twin, identifying up to 90% of potential issues before any physical modifications are made.

Initial outcomes: a 20% increase in throughput, a 10–15% reduction in capex, and nearly 100% design validation before a single physical change was made. Workers using the Siemens Industrial Copilot, which draws on the same knowledge layer, have seen a 30% productivity increase in early pilots, with 25% less reactive maintenance time.

The knowledge came first. The AI followed.

Read more: NVIDIA — Siemens Case Study

1) SAP Signavio Process Consulting Agent

A new AI agent embedded within SAP Signavio that provides process insights, reasoning, and recommendations in one interface. It connects process intelligence data with AI reasoning to surface where processes are underperforming and what to do about it — without needing to manually interpret the underlying data.

Read more: SAP Community — Process Consulting Agent

2) TAAFT: There’s an AI Tool for That

A searchable directory of AI tools organised by function and use case. For process and transformation teams evaluating vendor tooling, it provides a fast way to check what AI capabilities exist in a given area before beginning a formal procurement process.

Read more: There's An AI For That — AI Tools Directory

3) AI Readiness Assessment for Organisations — Udemy

A practical course covering how to assess organisational readiness for AI across people, process, and technology dimensions. Designed for team leads and transformation professionals who need to structure and communicate the readiness conversation internally.

Read more: Udemy — AI Readiness Assessment for Organisations

Samsung has got its eye on the ball

Ballie rolls around your home, projects images, answers questions, controls your smart devices, and can follow you from room to room. Samsung and Google Cloud confirmed the expanded partnership in May, embedding Gemini to give Ballie contextual awareness of your home environment and daily routines.

Ballie learns your routines, greets you in the morning, and apparently plays with your pets. The question everyone is asking is whether it is an AI assistant that moves, or a moving object that has AI. Either way, it has 10,000 pre-orders and is arriving in living rooms this year.

Read more: Samsung News — Ballie + Google Gemini

We talk to practitioners, leaders and specialists who are doing this work for real. No theory for the sake of it. Just honest conversations about building structured, data-led and outcome-focused approaches to change.

Listen here: bpmd — Work in Process Podcast