From Chaos to AI Orchestration

If you've ever worked in professional services or consulting, you knowthe dance: a promising client conversation wraps up, everyone's excited aboutthe project, and then... the waiting begins.

Waiting for the guidance document. Waiting for estimates. Waiting forapprovals from delivery, engineering, and finance. Waiting for someone toupdate the SOW. Waiting for everyone to find the latest version buried in anemail thread from three days ago.

For us, this wasn't just annoying—it was actively hurting our abilityto move fast. We know how to move quickly on complex technical work. But ourown internal process for getting from "yes, let's work together" to"here's the signed contract" was stuck in email archaeology andmanual handoffs.

So we built Lucia.

What Was the Real Problem?

On the surface, our challenge looked like a documentation problem. We needed to create three types of documents for every project:

• Guidance Documents - Initial estimates and requirements
• Estimate Documents - Detailed hour breakdowns, team composition, budget ranges
• Statement of Work (SOW) - The actual contract

But the real problem was coordination. Sales needed input fromdelivery. Delivery needed engineering's take on complexity. Finance needed toreview pricing. Everyone needed everyone else, and the process lived across Teamsmessages, email threads, shared drives, and "hey, did you see mymessage?" conversations.

Documents got lost. Context disappeared between handoffs. Versioncontrol meant adding "_v2_final_ACTUAL_final" to filenames. And worst of all, promising projects sat in limbo while we shuffled papers (or in this case, documents).

Why Did We Build Instead of Buy?

We looked at project management tools, workflow automation platforms, all the usual suspects. They could manage tasks, but they couldn't understand context. They couldn't read a rambling client email and know what questions to ask next. They couldn't maintain the thread of a project as it moved through multiple phases and departments.

We needed something that could actually orchestrate the workflow—not just track it.

We had a thought: what if an AI agent could act like a senior consultant who guides teams through this process, asks the right questions, keeps everyone aligned, and can produce valuable output?

That became Lucia.

What Changed?

The workflow itself looks similar on paper—we still create the same documents. But the experience is completely different.

Before: A salesperson would draft a guidance document, email it to delivery, wait for feedback, incorporate changes, email finance, wait again, lose track of which version was current, and eventually (maybe) get approvals.

After: Our team interacts with Lucia right where they alreadywork—in Microsoft Teams. But here's what's different: Lucia doesn't just coordinate—she actually produces the documents.

She asks the right questions, gathers the necessary information, and creates professional guidance documents, detailed estimates, and contract-ready SOWs that follow our standards. She's an active digital worker on our team, notjust a routing system.

Of course, there are human approval loops throughout—stakeholders review and approve at each phase. But Lucia is the one doing the work: synthesizing information, writing the documents, maintaining context as she moves from guidance to estimate to SOW. No more version confusion. No more lost approvals. No more "who's waiting on what?" And no more teams spending hours drafting documents manually.

What Did We Learn Building an AI Orchestrator?

Orchestration beats automation

We didn't just automate steps—we built a system that understands the workflow,  maintains context across phases, and coordinates between specialized functions. That distinction matters more than we expected.

Meet people where they work

Lucia lives in Teams because that's where our team already collaborates. We didn't make people go to a separate portal or learn a new interface. Adoptionhappened naturally because the friction disappeared.

Approval gates build trust

We don't auto-generate and send SOWs without human review. Lucia manages the workflow, but humans make the decisions. Those gates ensure quality and give stakeholders confidence in the system.

Your Turn!

If you're drowning in coordination overhead or manual handoffs—you're not alone. Everyone is facing some version of this challenge. The question isn't whether AI can help. It's how to apply it thoughtfully to your specific bottlenecks.

We learned a lot building Lucia—about multi-agent orchestration, state management, approval workflows, and what it actually takes to get teams to adopt AI tools. Some of it worked beautifully. Some of it we had to rethink entirely.

Want to hear more about how we built Lucia and what we learned along the way? Reach out—we're happy to share our findings with your team.