The what and when. The who and why.

The best technology handles routine coordination, improves the signal, and gives people more room for judgment, relationships, and purpose.

Fit the technology to the work

Development teams have no shortage of software. Email, schedules, budgets, pay applications, procurement logs, design platforms, dashboards, chat, and reporting tools all promise a cleaner process.

Yet a team can adopt more technology and still spend its day reconciling seventeen open tabs.

The problem is often fit. Off-the-shelf systems are built to serve a broad market. Development work is specific. The delivery model, approval path, cost structure, jurisdiction, stakeholders, and operating agreements can vary from one organization to the next and from one project to another.

When the system cannot carry that nuance, the team becomes the integration layer. Project managers copy information between tools, translate one format into another, chase confirmations, and explain exceptions that the system treats as errors.

Technology should remove that work, not make it more orderly.

Three teams, one recurring lesson

Across Tesla, WeWork, and Industrious, different teams approached the same challenge in different ways.

At Tesla, internal teams built focused tools around the work the retail development group actually needed to do. The value was not complexity. The tools did the necessary work and left out much of what the team did not need. Seeing that approach in practice made the impact of fit-for-purpose software clear.

At WeWork, experimentation was part of the culture. The pace created enormous pressure, but it also gave teams permission to improve the process while delivering the work. Our development team pushed the tools available to us as far as we could. We customized Smartsheet workflows, connected milestone reminders to Slack, automated repeatable updates, and worked closely with an embedded developer to keep refining the system around the needs of the team.

At Industrious, the business model created another kind of nuance. Management and revenue-share agreements meant that buildout costs and responsibilities varied by location. Standard accounting and construction platforms struggled with those terms. We evaluated several options, selected Banner, and formed a small working team with two project managers and Banner's developers. Together, we built a workflow that could reflect the deal structure of each project rather than forcing every project into the same template.

We also asked whether scheduling could live in the same environment. The intent was simple: reduce the number of places a project manager had to look before understanding the state of the work.

AI expands what the system can carry

Earlier automation depended on rules that had to be anticipated and written in advance. AI can now work with less structured information, identify patterns across documents, and surface exceptions that a conventional workflow may miss.

That creates practical opportunities across development and professional services.

An AI-supported first pass can review municipal and county procurement platforms against defined criteria, return a short digest, and help a team decide which opportunities deserve human attention. The result is not an automated bid decision. It is a better use of pursuit time.

Bid leveling is another example. A system can compare subcontractor or general contractor entries, identify missing scope, flag inconsistent assumptions, recalculate comparable totals, and show the estimator where the apparent low bid may not be the lowest complete bid.

First-pass estimating can draw on historical costs, prior estimates, labor assumptions, and industry data to create an informed starting point. A qualified person still reviews the output, challenges the assumptions, and owns the decision. The technology shortens the path to that higher-value work.

In strategy and operations work, AI can also help turn a large body of customer, market, or process information into usable signals. Our work with MAREA used that principle to help the team find insight that could inform the process rather than simply produce more reporting.

We have also collaborated on Electron PM around a long-held goal: one adaptable project environment that uses AI naturally, respects the nuance of the work, and reduces the need to assemble a project picture across disconnected systems.

Improve the signal-to-noise ratio

A useful technology strategy is not measured by the number of tools deployed. It is measured by the number of unnecessary decisions, handoffs, searches, and repeated explanations removed from the team's day.

The system should make the important signal easier to see. An exception that needs judgment. A risk that is becoming material. A decision that is holding three other decisions in place. A team member whose workload is about to become unreasonable.

Everything else should become quieter.

This is why technology should feel almost invisible when it is integrated well. People do not need another dashboard to manage. They need the right information to arrive when it changes the decision.

Start with the bottleneck

AI experimentation has real costs. Data has to be prepared and protected. Models consume computing resources. New workflows require testing, ownership, and maintenance. A clever demonstration is not the same as an operating improvement.

The right starting point is a specific bottleneck. Which repetitive work consumes capable people? Which handoff creates the most delay? Which analysis is valuable but too time-intensive to perform consistently? Which exception is discovered too late?

Then simplify the process before automating it. Build the lightest system that can create a meaningful improvement. Keep a qualified person in the decision. Measure whether the work became clearer, faster, or more reliable for the team.

The goal is not to put technology at the center of the work. It is to give the team more room to do the work only people can do.