Workforce agents

Let's start with workforce agents. Think about what your consultants do every day. Every customer needs their environment set up. Every customer needs roles and permissions configured.

Every customer needs workflows built, integrations connected, settings registered, and so on. They also validate what's been set up against what is promised, and this involves clicking through screens, checking boxes, catching mismatches before the go live. This work isn't creative. It isn't strategic, but you still have to do it, and this is where your team's time disappears.

A consultant might spend two days configuring something that looks almost identical to the last thirty configurations. Workforce agents now do this work for you. Set up customer environments according to your customer's needs, verify what's built, matches what is promised, catching gaps before the customer ever sees them. We want to give your consultants their own personal agent sidekicks so that they can get back to the work that actually require a human.

So let's take that exact work, configuration and validation, and see how it actually happens with an agent. At its core, your consultant is doing a few things. They take inputs and SOW requirements data from a customer, usually messy and unstructured. They structure it. They apply rules. They configure the system, and then validate everything before go live. We've broken this down into what we call skills.

Skills are reusable units of work. Things like extracting structured data, validating against your rules, or configuring your customer's account in your system.

These are the exact steps your team already performs.

Just package so the agent can now run them. Instead of stitching all of this together manually across tools and screens, you just describe what you want. Workforce force agents are designed as a completely conversational experience. So you tell the agent what needs to be done, and it figures out how to use these skills to get there.

Let's see it in action. Say Chargebee is implementing Amplitude and needs to configure their pricing plans in Amplitude's Chargebee instance. First, we define what Nitro needs to get started. In this case, the pricing pages address.

Next, we pick the skills the agent will use. One to extract pricing plans from the website, one to validate the extracted data's format, and one to push those plans directly into Chargebee. Then we connect it to Chargebee via their MCP. We'll describe our objective here in the builder, and that's it.

Nitro reads everything we've configured, our objective, our skills, our inputs, and checks that they all work together. It plays out the full plan, every step in order. We just have to review it and confirm. Once the agent is built, it lives right inside the project alongside your other agents and tasks.

The consultant doesn't have to go anywhere else.

We'll invoke the agent and type in the pricing page's address. Nitro gets to work. It goes to the pricing page, reads it, extracts the plans, and structures everything.

Then before configuring the plans in Chargebee, it shows the consultant exactly what it found. There is a human in the loop step for that added protection.

If something looks off, the consultant can adjust it right there. If it looks good, they go ahead.

Done. All four plans configured in Chargebee. The agent reports back with a full summary of what ran, what was created, and what to check next.

Switching over to Chargebee. The amplitude product family created four plans complete with entitlements and pricing details. For the consultant, it's none of the hassles of figuring out the pricing details, figuring out entitlements, and manually entering data. Hours of work reduced to a few minute.