The most valuable AI opportunity may exist between the systems you already own
Before replacing a platform, examine the work people do to bridge the gaps around it.
Good systems can still create fragmented work.
A business can have a capable CRM, a reliable accounting platform and a well-organised document store, yet still rely on people to join the dots. The information exists. The context needed to act on it is scattered.
Consider a routine customer decision. The commercial history sits in one system, the latest conversation in email and the financial position in another. A person assembles the picture, resolves inconsistencies and carries the decision back into the workflow. The opportunity sits in that repeated effort.
Look at the handover.
Before proposing a replacement, trace the points where work crosses a system boundary. What needs to move? What needs to be understood? What is copied manually? Which exceptions require judgement? These questions describe the business problem more clearly than a list of available integrations.
A connection alone is not enough. Moving information faster can amplify an existing ambiguity. The workflow needs a shared definition of the records, rules and responsibilities involved.
Connect context to action.
An intelligence and orchestration layer can be designed around a particular task: retrieve the permitted information, assemble relevant context, apply the business rules and prepare the next action. The person responsible can then review a meaningful result rather than reconstructing the situation from scratch.
Start with a contained use case. Establish which system remains authoritative, who is allowed to see the information and which actions require approval. Make exceptions visible and provide a practical route back to manual work.
From our work: a marketplace boundary.
2ndChannel keeps inventory and listing preparation in one workspace, while marketplace actions can run through the owner’s connected Mac. Where a suitable API is unavailable, browser automation works through the marketplace interface. This makes the handover between approval, execution and the returned status part of the product design.
A queued action and a completed marketplace change are different states. The operator needs to see the result or exception before relying on it. Connecting systems is useful only when the status of the work remains understandable.
Preserve what already works.
Replacement can be the right decision when an existing platform cannot support the required outcome. It should follow from evidence about the work and the constraints, rather than enthusiasm for a new technology.
The useful first question is whether the systems you already own could become more valuable by working together. A focused prototype can help answer it before a larger commitment is made.
Intelligence in practice.
