Agent Workplace.
Anywhere's first agent-facing CRM, shipped to MVP in six months after a two-year effort with significant investment had stalled. Built on AI from day one, designed to orchestrate over the patchwork of third-party tools agents were already using rather than force a migration.
// The situation
Two years in, with nothing to show.
When I joined Anywhere, the project had been running for two years with significant financial investment behind it. The infrastructure was largely built, but there was no design and no clear direction on what to actually ship.
The deeper problem was that Anywhere's six brokerage brands were each operating with a patchwork of third-party CRMs and productivity tools. The same capabilities had been bought from multiple vendors and bolted onto different brands. This was the first attempt to build the product internally and consolidate the work in-house.
// The team
Inherited a team that had never shipped from scratch.
I didn't get to hire. The team was already there: twenty product managers, engineers, and designers who had been working on this for two years. The honest read was that almost none of them had built a product from zero before. They were skilled at enhancing existing products, but launching a new one is a different kind of work, and they had stalled out trying.
The first move wasn't hiring. It was restructuring how the team operated: tighter ownership, clearer decision-making, and a discovery-to-delivery cadence that produced shippable artifacts every two weeks instead of two quarters. The cultural shift was as important as any roadmap decision.
// The bet
Orchestrate, don't rip and replace.
The fastest path to value wasn't forcing tens of thousands of agents to migrate off the tools they were already using. It was to build Agent Workplace as an orchestration layer sitting above the existing third-party products. Agents could keep their current tools while a single, brand-consistent UX absorbed them over time.
The second non-obvious bet: AI from day one, not bolted on later. Ava, the AI virtual assistant, was built into the foundational data layer. The system pulled data out of transaction documents and pre-populated databases to eliminate redundant data entry. It watched workflow patterns to suggest next-best actions. It was the orchestration engine that made the consolidation work, not a feature added at the end.
// Selling the vision
Six brokerage brands, one product.
The hardest part of the work wasn't technical. It was selling the vision to our broker partners. Each of the six brands had been used to making their own product decisions, picking their own vendors, and tailoring tools to their agents. Asking them to align on a single internal platform was asking them to give up control they had defended for years.
The work was relationship work: showing each brand what the orchestration approach meant for them specifically, how their existing tools fit into the new layer rather than being replaced, and what data and insights they would gain that their fragmented stack couldn't deliver. Buy-in came one brand at a time, and Coldwell Banker — the largest in the family — going first set the precedent for the rest.
// What we built
Agent Workplace.
An all-in-one productivity hub for real estate agents that brings lead management, transaction handling, marketing, and analytics into one platform across mobile and web. Used by agents across all six Anywhere brands.
Workflow as the spine.
Every interaction in Agent Workplace flows through Workstreams: the unit of work that follows a lead from first contact through transaction close. Agents see deals laid out as kanban columns by stage (Preparing, Marketing, Under contract, Recently closed). AI orchestration adapts the workflow to how each agent actually operates rather than forcing a single playbook.
Ava: the AI layer beneath everything.
Ava is more than a chatbot. It is the layer that reads transaction documents and pre-populates databases (eliminating one of the most-hated parts of an agent's day), watches activity patterns to suggest next-best actions, deflects high-volume L1 support questions, and acts as the front door to specialist support like Agent Concierge and Deal Advisor.
It is deployed as a standalone app inside Anywhere's Desk environment and integrated directly into Agent Workplace, so agents reach the same intelligence whether they start from a workstream, an inbox, or a question.
Ava and the broader AI work in Agent Workplace run on the Anywhere Intelligence Platform (AiP), the enterprise AI infrastructure layer we built in parallel to power AI products across the company.
Mobile-first because agents are field-first.
Real estate agents don't work at desks. The mobile experience launched alongside web because agents needed full workflow capability between showings, in cars, and at closing tables.
// What it produced
Six months from concept to MVP.
- 6 mo.
- Concept to MVP, after two years stuck.
- 7,000
- Agents in the initial test groups.
- 60%
- Agent return rate in early rollout.
- 30%
- Support cost reduction via Ava L1 deflection.
- $10M / yr
- Planned reduction in third-party software spend.
- Coldwell Banker
- First major brand live, the largest in the Anywhere family.
Beyond the metrics, Agent Workplace surfaced broker insights earlier in the transaction process than any of the displaced third-party tools could. That meant title and mortgage attachment opportunities became visible while agents could still act on them with clients, turning a productivity tool into a revenue lever for Anywhere's adjacent services businesses.
// What it proves
Scaling a product is rarely a technical problem.
Most of the infrastructure at Anywhere had already been built. The technology wasn't the bottleneck. What was missing was clarity (what are we actually shipping), alignment (which brand goes first and why), and a team operating model that could finish a thing.
What I took from this work: the hardest part of leading product in a large enterprise is rarely a technical decision. It is the human one, the political one, and the cultural one. Solve those and the technology follows. Anywhere had spent two years and significant investment proving the inverse.
// Keep going