Jeromee Johnson, President & CTO, Tellus App - Building the new and impactful at the intersection of PropTech and FinTech.getty​Most product teams are trained to see friction as the enemy.​That training is mostly right. A sign-up flow that requires nine fields when three would do is friction. So is the eighth password reset of the week, or a maintenance ticket that bounces across four systems before reaching anyone who can fix it. Discipline in removing friction builds better software. ​But not all friction is the same. The technology leaders being trained right now will spend their careers inside systems where the wrong removal of friction might move money, grant access, trigger trades, approve credit, dispatch vendors and ship code into production before anyone has stopped to ask why the old guardrail existed.​Three Kinds Of Friction​Three kinds of friction live inside most products. The first is waste—duplicate data entry, manual routing, status-chasing, disconnected systems. Remove it. The second is confusion—too many options, unclear next steps, hidden trade-offs. Redesign it. The third is protection—underwriting, confirmation, cooling-off periods, identity verification, human escalation. Be careful.​The third kind is the one that sometimes gets miscategorized. It can look like waste because it slows a user down. It can look like confusion when it adds steps. But its job is to give the user a moment they may need to understand the consequence before the system moves. That moment can look inefficient on a dashboard, but sometimes, it is the whole point.​Agentic AI Changes The Equation​Agentic AI makes the distinction urgent. McKinsey's March 2026 analysis of agentic AI in real estate directly frames the shift from "help me understand" to "help me get it done." The firm estimates the global productivity opportunity at $430 billion to $550 billion annually across real estate, construction and development. This same shift is unfolding across industries—financial services, healthcare, manufacturing—every domain where a multistep workflow can be compressed by an agent acting on the user's behalf.​McKinsey's most useful framing inside that piece is not the dollar figure. It is the distinction between automating steps and protecting thoughts. Automate the repeatable work and preserve human judgment for the exceptions, trade-offs, trust calls, financial risk and regulatory risk. The friction that should survive the automation pass is the friction that lives at the boundary between those two categories.​Automating Tasks Without Automating Judgment​NIST's AI Risk Management Framework points in the same direction from the governance side. Trustworthy AI has to be incorporated into the design, development, use and evaluation of the system. Governance has to sit inside the workflow. The risk is that the answer becomes a work order, payment, access change, customer communication or production update before anyone accountable has seen the decision. Protective friction is the control layer between model output and real-world action.​Where Protective Friction Matters Most​There are concrete versions of this that run across industries.​In maintenance and operations, the friction worth removing is the chase work: the status update that lives only in someone's inbox, or the ticket that has to be reopened because the first system did not talk to the second. Build the agent to do the diagnostic and present the choice. Do not build the agent to approve a $40,000 capital expense as if it were the answer to a $40 part.Software deployment has its own version. The friction worth removing is the manual release process where approvals happen through Slack archaeology, status lives in five dashboards, and nobody knows what is actually in production. Remove that, but do not remove staged rollouts, validation, rollback gates, separation of duties and human stop points for high-impact changes. The July 2024 CrowdStrike outage was a reminder that a small technical change, shipped at machine speed, can escape the screen and become an operating problem across the real world. The lesson is not “ship slower,” but “know which release frictions are protecting the system.”​Machine-Speed Decisions, Real-World Consequences​In consumer finance, Robinhood's $7.5 million 2024 Massachusetts settlement gives the same lesson from the other direction. Robinhood agreed to remove emojis from the transaction life cycle and permanently stop using confetti or other celebratory imagery directly tied to trading frequency, generalized push notifications highlighting specific lists, waitlist tapping features and features that mimic games of chance. The point was not that trading apps need to be hard to use. It was that speed, celebration and game mechanics can change the decision the user thinks they are making. The friction that had been removed was doing protective work.​In real estate, some friction is real and some is just legacy. A house does not transact slowly because slow is noble. It transacts slowly because the information needed to move it safely is distributed, asymmetric and hard to verify: title, liens, condition, rent history, insurance, appraisal, local rules, inspection findings and cash flow. There is a real technology opportunity to remove the operational friction—the document chase, the manual reconciliation, the rent-verification cycle that takes 30 days because the data lives in 15 places—without turning the asset into something that trades in seconds. A 30-day close that becomes a six-day close is progress, but a 30-day close that becomes a 30-second close is something else.​Before You Remove Friction, Asks Why It Exists​The practical standard for technology leaders is not complicated.​Before removing friction, ask what it was doing. If it was protecting the user from a decision they cannot easily unwind, replace it only with a cleaner safeguard: a threshold, confirmation, staged rollout, approval path, cooling-off period or a human stop point.​The goal of good technology is not to make every action instant. It is to make the right action easier​. Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?