What does a custom AI build actually involve?
A custom AI build starts with a model call and a prompt, which takes an afternoon. A production agent that a business can rely on takes much longer, because it needs everything around the call: a way to propose rather than act, a human approval step, an audit trail of decisions, secure handling of the credentials it uses, isolation between the properties it works on, metering and limits on spend, a memory of the business that a person can inspect and correct, and an interface where a manager can review all of it.
Most of that is not AI work; it is product and platform work, and it has to be maintained as models, tools, and requirements change. The build-versus-buy guide on ftes.ai walks through the list in detail.
What does a Digital FTE deliver in week one?
A Digital FTE delivers that layer already built: the approval gate and autonomy ladder, the audit trail, encrypted scoped credentials, organization and property isolation, the token wallet with a hard stop, the Brand Brain memory with provenance, and the manager workspace with its morning brief, approvals, pipeline, targets, performance, missions, memory, and setup tabs.
It also delivers a role-specific remit, such as the fact gate and worth-it verdict for a Search Visibility FTE or the evidence-per-match behaviour of a Finance Controller FTE, which is the part of a custom build that takes the longest to get right.
Where does a custom build win?
A custom build wins when the agent is your product or your core differentiator: a capability your customers buy, or a workflow so specific to your business that no general role fits. It also wins when you have the engineering team to build and maintain the governance layer and the appetite to own it.
Control is real. A custom build can do exactly what you specify, integrate with anything you can reach, and change whenever you decide. The cost of that control is that every part of it is yours to maintain, including the parts that have nothing to do with your differentiation.
- Build: the agent is your product or core differentiator
- Build: the workflow is unique and no general role fits
- Hire: the function is standard, such as reconciliation, screening, support, content
- Hire: you want the governance layer without owning its maintenance
- Both: build the differentiating capability, hire the standard functions
What does maintenance look like on each path?
On the custom path, maintenance is continuous: model updates change behaviour, tool APIs change, security requirements evolve, and every improvement to the governance layer is your engineering time. An agent built in year one is a system you support in year three.
On the Digital FTE path, the platform maintains the layer, and your maintenance is management: refining the remit from send-backs, confirming the memory quarterly, deciding autonomy promotions, and reviewing performance. That is manager time rather than engineering time, which is usually the scarcer resource for the functions in question.
How should you decide?
Decide by asking whether the function is something you sell or something you need done. If a customer pays for it, or it is the thing that makes you different, build it and own it. If it is a business function that every organization needs and yours needs done well, hire a Digital FTE and manage it.
The two paths coexist well. Organizations that build their own differentiating agents still have reconciliation, screening, support, and content to do, and those functions are where a Digital FTE fits without competing with what the engineering team is building.
How do a built agent and a Digital FTE coexist in one organization?
They coexist by dividing the work along the line of differentiation. The engineering team builds and owns the agent that is the product or the core capability, using whatever framework and provider it chooses, and owns the governance layer around it because that layer is part of the product. The rest of the organization hires Digital FTEs for the standard functions, reconciliation, screening, support, content, and search visibility, and manages them through the workspace with the governance layer already built and maintained by the platform.
Where the two need to talk, a Digital FTE on ftes.ai can be exposed to an external assistant or agent through a connector that carries its credential once and is rotated rather than recovered if lost, so the organization's own agent can query the FTE's work while the FTE's approval boundaries stay in place. The engineering team is usually glad of the arrangement, because reconciliation and support were never the work they wanted to own, and the finance leader is glad because the standard functions sit on a metered line with a hard stop rather than on an engineering roadmap.