A different model for building AI agents
Agent Factory
Most AI consulting starts with a project, a statement of work and a development bill. Agent Factory works differently.
Bring us a business process. We’ll determine whether an AI agent can realistically perform it, what that capability could be worth to your business and what it should cost to operate. If the opportunity makes sense for both of us, Forceworks funds the development.
Your environment
Your Azure. Your control.
Your agents run on RapidClaw in your Microsoft Azure environment—not ours.
Your cloud resources remain under your control. You pay Microsoft and AI model consumption directly, and your core agent infrastructure stays within your Microsoft estate except where the agreed process requires connections to external systems.
From experiments to operations
Close the gap between trying AI and changing the business
Most businesses aren’t starting from zero. You’ve used ChatGPT. Someone has experimented with Copilot. Maybe you’ve built prompts, tested an automation or launched an internal pilot.
And yet very little about the way the business actually operates has changed.
You bring the process knowledge
You shouldn’t have to build an internal AI engineering organization just to get meaningful value from AI. You know the processes that run your business.
We turn the right process into an agent
We don’t begin by selling a prebuilt Sales Agent, Finance Agent, Service Agent or HR Agent. We start with the work your business is doing today that an agent might be able to do instead.
Start with the process
One worthwhile process. One focused workshop.
Maybe the work means researching across systems, reviewing incoming requests, monitoring continuously, applying business rules, preparing a response or taking action. Maybe it consumes a day of someone’s time every week.
Agent Factory starts with the process, not the technology.
01
Workshop
Map the process, people, systems, frequency, time and cost.
02
Underwrite
Assess business value, feasibility, consumption, dependencies and risk.
03
Build
Forceworks funds development after the business case and criteria are agreed.
04
Prove
Test the agent against the success criteria set before development.
05
Operate
Begin the subscription only when the agent meets those criteria.
Underwriting
Not every process should become an agent.
We look at likely business value, technical feasibility, expected AI consumption, dependencies, integration requirements and development risk.
Because we invest our own development resources, we have just as much incentive as you do to avoid agents with weak economics. If the business case doesn’t work for both of us, we’ll tell you. No project gets created just because someone has budget available.
Reasons we may recommend not proceeding
- ✓The process does not create enough value
- ✓The underlying technology is not reliable enough yet
- ✓AI consumption would cost too much
- ✓Critical data is not accessible
- ✓The process changes too frequently
- ✓A required integration is disproportionately expensive
- ✓Development risk exceeds the value of the opportunity
Clear acceptance criteria
Specific capabilities, agreed before we build
Not vague promises about “AI transformation.” A defined process, performed to an agreed standard.
We also define an agreed level of accuracy, reliability or throughput—and the monthly subscription that will begin if the agent meets those criteria. You know the price before we build.
We take the development risk
We don’t get paid for trying.
There is no development invoice. You’re not paying our engineers while we determine whether the idea works, financing our experiments or buying a pile of hours and hoping there’s something useful at the end.
When the agent is ready, we test it against the criteria agreed before development. If it hasn’t met them, the subscription doesn’t start. If it has, the agent goes into operation and your subscription begins.
A managed capability
Built to evolve with AI
Models change. Capabilities improve. Frameworks evolve. Your subscription includes maintaining the agreed agent capability and adapting it as the underlying technology changes.
If we can make the agent faster, more reliable or less expensive without changing its agreed purpose, that’s part of the service. A completely new capability is different—that’s another opportunity, and possibly your next agent.
Accountability
Pay for a working capability—not development effort
Traditional consulting charges for effort. Pure outcome-based consulting can make a provider responsible for business results shaped by factors outside its control. We use a better boundary.
Forceworks is accountable for
Delivering an agent that performs the defined process to the agreed standard, then maintaining that capability as part of the subscription.
Your organization controls
What you do with the time, capacity or opportunity the agent creates. That keeps accountability where it belongs.
Value-based subscription
The price reflects the capability’s value—not our hours
A simple agent can automate an expensive process. A difficult agent can solve a problem that isn’t worth much. Development effort and business value are not the same thing.
We estimate value from factors such as:
- ✓Labor capacity returned to the business
- ✓Increased throughput
- ✓Faster cycle times
- ✓Reduced operating costs
- ✓Avoided costs
- ✓Increased revenue capacity
- ✓Improved utilization of existing staff
We then account for expected operating costs, including AI model consumption. The subscription is agreed from the potential net value before we build.
Illustrative monthly economics
Every process will have different economics. The principle is constant: there must be enough value for both sides to win, and Forceworks funds the development.
Aligned economics
The Agent Factory only works when the agents work
If we think an agent can create enough value, we invest our resources to build it. If we’re wrong, that’s our development risk. If we’re right, we earn a recurring subscription for maintaining a valuable capability.
You don’t want marginal agents.
Neither do we.
You don’t want AI experiments that go nowhere.
Neither do we.
You want a capability that keeps working.
That’s what the subscription maintains.
RapidClaw in your Azure
No Forceworks data silo. No token markup.
Agent Factory agents run on RapidClaw, our business agent platform built on OpenClaw and deployed into your Microsoft Azure environment—not a Forceworks SaaS environment.
You pay Microsoft and AI model consumption directly. The agent can use your existing Microsoft identity, security, governance and infrastructure investments while connecting to external systems where the agreed process requires it.
Continuous optimization
We don’t benefit from wasteful AI consumption.
Operating cost is part of the business case before we agree to build. If a better or more specialized model can deliver the agreed capability efficiently, we can adopt it as part of the managed service. Our subscription does not increase because the agent becomes more efficient.
If third-party AI costs materially increase, we first seek optimization, different models or other technical approaches. If the original economics cannot reasonably be restored, we review the agent with you rather than trap you in a business case that has disappeared.
A factory model
Grow process by process—not through one enormous transformation project
Find a process. Underwrite it. Build the agent. Prove it. Operate it. Then find the next one.
Every qualifying agent is built with no development fee—not just your first one.
What makes a good candidate?
We look for processes frequent, valuable and structured enough to justify an agent.
- ✓Repetitive knowledge work
- ✓Multi-step processes
- ✓Information gathering and synthesis
- ✓Decision rules and routing
- ✓Interaction across business systems
- ✓Work that consumes significant employee time
It doesn’t need to replace a job.
Sometimes the benefit is reduced labor cost. Sometimes the same people can accomplish more valuable work. Sometimes the benefit is speed, responsiveness or throughput.
We’re interested in the economic value of the process—not your headcount strategy.
Infrastructure dependencies
An agent needs access to the systems and information it uses.
Some processes can use capabilities that already exist. Others require an API, connector, MCP server or data layer. We identify those dependencies before committing to the build.
We may decide that creating required infrastructure belongs in our investment. In other cases, it may need separate funding—or make the overall opportunity unattractive. We fund the agent build; we do not automatically fund every unrelated system change needed to make a customer’s environment agent-ready.
Start with one process
Apply for a free Agent Workshop
This is a focused assessment—not an open-ended AI brainstorming session. We’ll understand the process, assess whether it is a realistic agent candidate and determine whether there is a business case worth pursuing.
Tell us about your organization, your role and the process you have in mind. If it looks like a reasonable Agent Factory candidate, we’ll send you a link to schedule your workshop.