There are three honest paths for one existing workflow. Build it with your own engineers. Turn on a product you already pay for, or buy a new one. Or have someone put a supervised agent live in the systems you already run, then keep a person on the writes that can hurt you.
Each path fails in a different way. The useful question is which failure you can live with, not which slide has the lowest number.
Building it
Your team already knows the process. They have credentials. They will still underestimate the same four things every internal build underestimates: the permission set, the exception owner, the colliding automations already in the tenant, and the monthly change work when Salesforce, Zendesk, or NetSuite ships a release.
An internal build is the right answer when you have an engineer who can own this after go-live, a security review that will happen before the agent writes, and a process owner who will still be in the chair in month four. If any of those is missing, you are staffing a pilot, not a run.
We do not tell a capable internal team to outsource a workflow they can finish. If that is you, stop here. Use the rest of this article as a checklist for your own build.
Buying a product
Platform vendors are shipping the easy half of this work. Zendesk has AI agents that can set ticket fields. Salesforce has Agentforce and Apex actions. NetSuite’s Bill Capture now reads invoices with a generative model. If the workflow is standard, lives in one of those products, and the native feature already covers the action you need, turn that on first.
Buying fails when the demo is a chat in the vendor’s cloud and production is a write into your CRM. It also fails when the product can extract or classify, and you still need a person to own the items that do not match a purchase order, a policy, or a routing rule. That gap is not a missing feature. It is operating work.
A branded platform is also the right answer if what you actually want is a product to rent, with their UI, their tenant, and their roadmap. We are the wrong supplier for that. We do not sell a platform.
Having someone put it live
That is the package on our homepage. One workflow. $9,900 to go live in the tools you already run. $2,900 a month to run it, with a named owner from our employed team. A person on the writes that can change money or a customer. Leave any month after it is live.
The failure mode is scope. If you need six systems, unsupervised replies to customers, and a transformation programme, this number is not a bargain. It is the wrong product. We will say that on the call.
It is also the wrong product if a native feature you already pay for will do the job after a week of configuration. We would rather send you back to your own admin than invoice you for wrapping a checkbox.
How to choose without a scoring matrix
If the workflow is standard and the platform you already pay for has a switch for it, buy that. If you have the engineers and they will still own it in six months, build it. If you need it live this quarter, in your tenant, with someone awake when an exception ages, that is the case for a partner on one workflow.
The assessment is free and 30 minutes long. Bring the process. If we are the wrong path, that is a useful half hour.