The Shift From Writing To Defining

For years, the center of gravity in software was implementation. We admired the engineer who could take an idea and turn it into working code. This will change in the near future.

My view is simple: If a problem can be scored reliably, autonomous agents will eventually learn to solve it. They will write, run, measure, revise and self-improve against that score faster than human teams can manually iterate. The human job does not disappear. It moves up a level. We still choose the destination. We still decide what trade-offs matter. But once “right” is defined, implementation becomes increasingly automatic.

This is why AI does not just automate coding. It exposes organizational ambiguity. Human teams can survive vague goals because experienced people fill in the gaps with intuition, memory and hallway conversations. Agents cannot. A vague company will get vague AI. AI is not only a coding tool. It is a mirror. It shows an organization how clearly it actually thinks.

Rightware Turns Wishes Into Scoreboards

Most business goals arrive as aspirations, such as improving onboarding, modernizing billing or using AI in customer support. Those are directions, not definitions.

Rightware turns those wishes into scoreboards.

Take onboarding. A vague request might sound like this: “Make the sign-up flow better.” A rightware version sounds very different: Increase activation by 15%, keep average setup time under five minutes, preserve accessibility standards and do not increase support tickets.

Take billing. “Rebuild the platform” is fuzzy. Rightware would say: Reproduce one year of historical invoices within an agreed tolerance, preserve every regional tax rule, finish month-end reconciliation on time and produce an audit trail for every exception. That is no longer a brainstorm. It is a road a machine can drive.

Take customer support. “Build an AI assistant” is not a plan. Rightware would define a benchmark of past cases, set a target for correct routing, forbid invented policies, require cited answers and force escalation when confidence falls below a threshold. Now the system can take careful laps around the track: answer, test, compare, revise.

This is the new craft. Not asking the machine once, but designing the conditions under which it can keep getting better. Rightware is judgment made executable.

Hiring And Team Design Will Change

Once you see this, hiring looks different. I would hire less for keyboard fluency and more for judgment. Can this person surface hidden assumptions? Can they identify the failure that would quietly hurt customers? Can they design an evaluation that reflects the real world instead of the demo? Can they tell the difference between a metric that is easy to measure and one that actually matters? And most importantly, can they turn a hard business problem into a verifiable, well-defined task?

The best engineers in the rightware era will not simply be great coders. They will be the people who can turn ambiguity into criteria, criteria into tests and tests into learning loops.

Team structure changes, too. Quality assurance stops being the last checkpoint and becomes a core engineering function early in the process. Product managers, designers, operators, security leaders and compliance teams need to be involved earlier, because rightware is where institutional judgment gets encoded. It is where the company decides, explicitly, what it means to be fast, safe, fair, compliant and useful.

Build The Roads Before You Buy Faster Cars

Many leaders are asking which AI coding tools they should adopt. I think that is the wrong first question. Most companies assume they have a model gap. Many actually have a clarity gap. The better question is this: Where, exactly, do we know what correct looks like?

Start there. Find the workflows that matter most and turn tribal knowledge into tests. Build benchmark datasets from historical decisions. Create sandbox environments where agents can change code without touching customers. Replay real edge cases. Make rollback fast. Reward teams not just for shipping features, but for increasing the share of the business that is covered by rightware.

This investment compounds. A stronger model only helps if your organization can better recognize the difference between good and bad. With strong rightware, each new generation of agents becomes more valuable. Without it, even impressive agents are just very fast guessers.

The future of engineering is not less human. It is more explicit. More intentional. More honest about what the business truly values. Code will become abundant. Judgment will not. The companies that come out ahead will focus less on sheer volume of code and more on clarity—defining what “right” looks like and creating the pathways that guide autonomous systems toward it.​

Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?