Manifesto — Performance Engineering Lab
September 2026 · 4 min
00 — The situation
We are seeing the biggest transformation of work in companies since the invention of the typewriter. The situation is completely quantum right now: happening and not happening at the same time.
Agents will finally deliver what the computer promised decades ago: freeing humans from mundane tasks. So far there is little to prove the point — the reality inside enterprises looks very different. Here is why.
01 — Humans as APIs
The enterprise runs on a handful of core systems: email, ERP, CRM, DMS, accounting. To this day they are complete data and action silos — running in parallel, mostly unconnected, unaware of each other.
This should have been a huge problem. Instead, enterprises deployed humans as APIs and context engines to make up for the architectural shortcoming.
Companies are paying twice: once for the software license — and once for the human who operates the software.
02 — The chatbot phase
The technology — LLMs, cloud infrastructure, software — is ready and could, in theory, run all of a company's operations autonomously. Yet adoption is slow: most enterprises are still in the chatbot phase, just starting to give employees access to LLMs via chat.
That augments people — it will bring none of the productivity gains agents will. It even cements how today's systems and processes are built: entirely around the human who initiates the actions, fills the gaps, copies from one system to another and keeps half the context in their head.
Continuing like that leads to exactly one place: total delusion about AI. The reports of failed enterprise AI initiatives already exist — along with CFOs complaining about absurd token costs.
03 — An architectural question
We have seen it first-hand, with ourselves and with our customers: agents bring teams full autonomy — and, more importantly, real-time intelligence about the business.
Knowing that, we believe we will see a Cambrian explosion of agentic actions in enterprises. It will make companies faster, more intelligent and far more competitive.
04 — Our hypotheses
05 — A holistic solution
The enterprise is more than the ERP, the CRM or the processes. It is everything together: emails, documents, ERP, CRM, HR, DMS, people …
We are building the Enterprise Context Protocol (ECP) to help enterprises deploy, operate and scale agents as fast and as easily as possible. Our goal: set up and run productive agents within hours.
06 — The Enterprise Context Protocol
The ECP is how any system — an AI assistant, a CRM, a BI tool — connects to a company's entire business through one connection, without knowing a single underlying API.
Sophia sits in the middle as the Context & Action Broker: a canonical graph of the enterprise — one node per real customer, order or employee, identity-mapped across every source. Every read runs through a handful of self-describing verbs. Every write becomes a governed, evidenced action, routed to the system that owns it.
Because Sophia holds all data from all connected systems in its own database, the limits of the underlying software stop mattering: a slow ERP syncs in the background at its own pace — while everything critical is escalated to a human, who approves the steps that matter with one click.
The result is real-time intelligence and execution: evidence arrives, Sophia acts in the systems of record, traces the impact and drafts the consequence — one click from done. And because every newly connected system reaches all the others instantly, the hub compounds: n connections replace the n² point-to-point integrations the Mittelstand pays for today.
The people stay exactly where they are irreplaceable — on judgment, exceptions and relationships — while Sophia does the part of their job that was only ever copying context between systems.
Freed from the constraints of legacy systems, enterprises will rapidly scale the deployment of agents — and become: faster, stronger, better.
Simon Tüchelmann · Adam Touzani · Seung-Zin Nam
Sophia