The BPM Market This Week: Convergence Wins, and "Context" Becomes a Contested Asset

A weekly read for BPM, Process Mining and Business Transformation practitioners. Week of 22 to 28 September 2026.

This week the loudest theme was not a new tool or a new number. It was convergence: BPM, Enterprise Architecture and Process Intelligence stopped being three conversations and became one. And underneath it, a quieter but sharper fight broke out over who gets to own the "context" that AI runs on. Here is the practitioner's cut.

1. Convergence stopped being a slide and became the buying conversation

Signals in the BPM + EA + Process Intelligence cluster jumped by roughly 60 percent week on week, the sharpest rise of any theme. Three pieces made the same argument from different angles: that deploying AI copilots tool-by-tool across process excellence, GRC, mining and architecture does not fix fragmentation, it accelerates it. A copilot for every silo is still a silo, just faster. The proposed fix is consistent: a shared semantic reference layer, in place before you scale AI augmentation.

For practitioners: if you are running several AI pilots across disconnected tools, the integration gap is now your bottleneck, not your model quality. The prerequisite work is process and architecture work, and it comes first.

2. Celonis was everywhere, which cuts both ways

The dominant vendor story came from one name in four moves: a Celosphere 2026 event (Munich, 2 to 3 December) built around "RoAI" - Return on AI; a new Field CTO hired specifically from an Enterprise Architecture background; a Context Model framed as "real-time process grounding for AI"; and continued consolidation of mining plus AI-driven modelling through acquisition.

The message - AI needs process context to pay off - is one we agree with completely. But the same week carried the counter-story: a lawsuit over ERP data access, alleging one vendor is using its ecosystem to lock out third-party process-mining tools. Context, it turns out, is becoming a contested asset.

For practitioners: treat the validation and the warning as one lesson. Yes, AI needs a source of truth about how your business actually works. No, you should not let any single vendor own that truth. The durable posture is tool independence: keep your process knowledge in a neutral, living model and choose the mining, modelling and orchestration tools that fit each job.

3. Agentic AI: govern the process before you scale the agent

A high-signal roadshow this week (with practitioners from Bentley, Airbus, Fraport and Suva) landed on one point: process understanding is a prerequisite for safe AI deployment, not an afterthought. Digital twins and guardrails are core infrastructure; human judgment stays embedded in the decision logic. Market forecasts kept the excitement high (agentic and AI automation projected into the trillions by the mid-2030s), and orchestration platforms pushed model-context and reasoning-engine capabilities.

The most useful reminder was blunt: when one enterprise tripled its automation volume, its exception volume tripled too. Automate a broken process and you generate broken outcomes faster.

For practitioners: design for "judgment-at-exception". Let agents run the flow; put people on the loop precisely where judgment matters. That only works if you genuinely understand the flow first.

4. And a new kind of competitor

A major systems integrator invested in a proprietary AI platform for process mapping and agentic automation this week, aimed straight at the process-intelligence space. Expect more large integrators to bundle their own toolchains. The differentiator that matters is the same as above: framework-led, vendor-neutral transformation versus a locked stack.

The bpExperts read

None of this is a technology problem you can buy your way out of. It is a process and organisation problem that technology can finally help you solve, if the foundation is right. This week the market confirmed our stance, loudly:

• Make process knowledge shared. One living model your teams and your AI agents both start from, owned by you.
• Put AI to work, safely. Agent-assisted execution with people on the loop where it matters, and governance built in from day one.
• Make change a capability. Integrated across the tools you already run, so you are never locked into one vendor's version of the truth.

Convergence was always coming. The open question this week was who owns the context it runs on. Our answer: you should.

Want the full weekly briefing, including the tenders and thought-leader signals behind this read? Talk to your bpExperts contact.