# A pre-built agent runs someone else's operation

> IBM and Microsoft now offer 6,000+ pre-built agents and assets for industry processes. A catalog is a start. What makes you different is still yours to design.

- Type: Opinion
- Published: 2026-11-11
- Author: Sentynel
- URL: https://sentynel.ai/sygnals/a-pre-built-agent-runs-someone-elses-operation/
- Language: en

IBM and Microsoft announced Imagine Enterprise on 10 November: a platform for building, running and governing agents on industry-specific processes, with more than 6,000 pre-built agents and assets from IBM Consulting, starting with financial services, insurance and retail. The easy reading is that the design work is now done for you: pick the agents for your industry and switch them on. **A pre-built agent encodes how a typical company in your industry works, and the parts of your operation that are not typical are exactly where your margin lives.**

## What a catalog actually gives you

A large library of industry agents is useful. It saves you from writing a claims-intake agent or a reconciliation agent from a blank page, and it comes with governance, identity and observability already wired in. IBM's own announcement says clients can adapt the agents and connect them to existing data, processes and systems. That word, adapt, is where the real project sits.

An industry agent is built from the common version of a process: the steps most insurers follow, the checks most retailers run. It is a good average. Your company is not the average, and the difference is rarely in the common steps. It is in the exceptions you accept, the customers you treat differently, the approval you never skip, the shortcut that only works because of how your team is organized.

## Adapting is designing, under another name

When a pre-built agent meets a real operation, one of two things happens.

- **The agent bends to the operation.** Someone has to know the operation well enough to say where the agent's steps are wrong, which rules are missing, and who takes the case when it does not fit. That is operation design.
- **The operation bends to the agent.** Nobody does that work, so the company quietly adopts the vendor's version of the process. It goes live faster. It also gives up whatever made the old process better than the industry's.

The second path looks cheaper because the cost does not show up as a line item. It shows up later as customers who notice you now behave like everyone else, and as staff who keep a spreadsheet on the side to handle what the agent cannot.

## Where pre-built fits, and where it does not

The useful question is not "build or buy". It is which steps are common and which are yours.

1. **Common steps:** document intake, data extraction, standard checks, routine notifications. Pre-built agents are a sensible default here. Being average costs you nothing.
2. **Differentiating steps:** pricing exceptions, how you prioritize a key account, how you decide what a campaign should be. These need to be written down from your own operation before any agent touches them.
3. **Handoffs between them:** who owns the case when it moves from a common step to a differentiating one. A catalog cannot know your org chart.

**Sorting an operation before choosing agents**

1. Map the real flow — As it runs, not as documented
2. Mark common steps — Same as most of your industry
3. Mark your steps — Where you win or lose
4. Assign agents — Pre-built for common, designed for yours
5. Name the owners — One per handoff

## A case where the advantage was the question

In the Matosantos / Campoverde case, sales and analytics became a shared decision layer, then a Walmart campaign was built from what the data revealed. Sentynel's case record reports a +18% sales uplift. Sales uplift is a reported campaign outcome; the chart uses a normalized baseline.

A generic consumer-goods agent could have produced a sales report. What mattered was that the board got a way to ask its own questions of its own data, and the campaign came from the answers. That part did not exist in any template, because it came from how that company decides.

**Matosantos / Campoverde · Walmart campaign**

- +18%: reported sales uplift
- 24: states coordinated

Source: Source: Sentynel's Matosantos / Campoverde case record. Sales uplift is a reported campaign outcome; the chart uses a normalized baseline.

## Use the catalog, keep the design

None of this is an argument against platforms like this one. Pre-built agents with governance built in will save real time on the steps that should be standard. The mistake is treating the catalog as the plan.

**Decide what your operation is before you choose which agents run inside it.** Then the pre-built agents take the common work, and the work that makes you different gets agents designed around it.

A test before any rollout: list the five decisions in one process that your best person would make differently from a competitor. If the pre-built agent you are about to deploy makes those five the way the competitor would, what exactly are you automating?

## Sources

- https://newsroom.ibm.com/2026-11-10-ibm-and-microsoft-announce-a-new-ai-agent-platform-designed-for-industry-specific-transformation
- https://sentynel.ai/#cases
