From cost center to growth engine

the situation

REDD Platform aggregates commercial real estate data from over 6,000 office and industrial assets across CEE - availability, lease terms, market intelligence. The business model is SaaS: high margins, scalable growth, costs that stay flat as revenue expands.

There was one exception. Data collection.

Landlords don’t use APIs. They call. They email spreadsheets. They share PDFs. For years, REDD operated a team that was part data-entry, part call center - receiving information however it came in and manually processing it into the database. The team was competent and well-tooled. Nevertheless, it was hardly scalable.

At peak intensity, 15-20 people were dedicated to this process. It was the single largest component of COGS. And every time REDD considered geographic expansion, the math was the same: new market means new landlords means more headcount - just to maintain the input, before a single euro of new revenue.

The data collection process was the major constraint.

the constraint

The core problem wasn’t efficiency. The team was already running lean. The problem was structural: the process was built around human interpretation of unstructured inputs - phone calls, email threads, inconsistently formatted spreadsheets. No two landlords communicated the same way. Standardizing that at scale required people.

Until generative AI changed what was possible.

When large language models matured to the point of reliably reading, transforming, and structuring unformatted data, the last bottleneck became solvable. What previously required manual copy-paste - interpreting a landlord’s spreadsheet, extracting the relevant fields, updating the database - could now be pushed through an automated pipeline.

The technical unlock was real. But the organizational readiness to use it wasn’t automatic.

the approach

Before building anything, REDD addressed the foundation: how the organization thinks about AI.

A company-wide AI strategy was written and presented internally. It established three non-negotiable principles:

First, AI skills are an individual responsibility - developed through use, not workshops. No expensive training programs. Everyone starts at the same point because the technology is new for everyone. Learning happens by doing.

Second, every team member is expected to build. Not consume AI tools - build them. Within three months, nearly everyone at the company had created at least one AI-powered tool tailored to their own workflow. The learning curve became practical, fast, and self-reinforcing.

Third, REDD committed to never hiring for roles that AI could perform at a comparable level. This raised the bar on what “necessary headcount” means - and focused the organization on leverage, not addition.

With this foundation in place, the data collection project became a company-wide quarterly priority. Research and IT teams designed, built, and tested the new pipeline collaboratively (using AI throughout the process itself!). The mindset shift made the execution possible.

the result

Emails from landlords are now automatically transformed into standardized, database-ready formats. The system compares incoming data against existing records and updates the database without human intervention when changes are detected.

Human involvement is now limited to quality control.

A process that required up to 20 people at peak is being restructured around a fraction of that headcount - with the remainder redeployed to business expansion. The prior cost center is becoming a growth engine: margin improves, the team gains higher-value work, and geographic expansion no longer requires proportional headcount growth.

what made it work

Two decisions mattered most.

The first was sequencing. The data pipeline was not the starting point - the AI strategy was. Building organizational readiness before tackling the hardest process meant the team approached the project with capability and conviction, not resistance. The foundation made the execution fast.

The second was the principle of never hiring for AI-replaceable roles. It sounds like a cost decision. It’s actually a culture decision. It creates permanent pressure to find leverage before adding headcount - and that pressure drives better outcomes than any top-down automation mandate.

relevance beyond REDD

Commercial real estate operates in a geography-first model. Brokerage firms, investment funds, asset managers, developers - all expand by entering new markets, and all face the same scaling constraint: adding territory means adding people to maintain local data, relationships, and processes.

AI-first operating models break that constraint. The same team can cover more ground. The same processes can handle higher volume. Expansion becomes a systems problem, not a headcount problem.

That’s the shift. And it starts with how the organization is built - not what software it buys.

When not running REDD, I work with real estate firms on exactly this - rebuilding operating models around AI, not buying more tools. 3-12 month engagements, 1-2 clients at a time. Read more