Rajneesh Kambhatla — essaysThe practice layer above the platforms. Strategy, agentic governance, enablement for Australian enterprises and scaleups.https://rajneeshkambhatla.com/en-AUThe prompt is not the programhttps://rajneeshkambhatla.com/essays/2026-07-04-the-prompt-is-not-the-program/https://rajneeshkambhatla.com/essays/2026-07-04-the-prompt-is-not-the-program/A prompt is configuration written in natural language — unversioned, untested, and one model update from meaning something else. Why the serious end of the industry is moving the logic out of the sentence and into the architecture.Fri, 03 Jul 2026 20:00:00 GMTD6 — accountability before the agent runshttps://rajneeshkambhatla.com/essays/2026-07-03-d6-decision-rights/https://rajneeshkambhatla.com/essays/2026-07-03-d6-decision-rights/Decision rights make autonomous AI governable by showing where authority, pause power, redress, and outcome accountability are missing before the system acts.Thu, 02 Jul 2026 20:00:00 GMTD5 — the shape of permissionhttps://rajneeshkambhatla.com/essays/2026-06-30-d5-agent-architecture/https://rajneeshkambhatla.com/essays/2026-06-30-d5-agent-architecture/Agent architecture becomes governable when every agent's autonomy, blast radius, escalation path, and recovery design are explicit before deployment.Mon, 29 Jun 2026 20:00:00 GMTThe illusion of a second opinionhttps://rajneeshkambhatla.com/essays/2026-06-29-illusion-of-a-second-opinion/https://rajneeshkambhatla.com/essays/2026-06-29-illusion-of-a-second-opinion/I stopped using AI as one assistant and started running it as a small team — a driver, an adversary, a reader, each a different model family with a job. The framework, and why a real second opinion is something you design, not something you default into.Sun, 28 Jun 2026 20:00:00 GMTThe tax on every other languagehttps://rajneeshkambhatla.com/essays/2026-06-27-tax-on-every-other-language/https://rajneeshkambhatla.com/essays/2026-06-27-tax-on-every-other-language/Why large language models charge most of the world four times over to read its own languages — how the bias gets built in at the tokenizer, and how a handful of teams across the Global South are repealing it.Fri, 26 Jun 2026 20:00:00 GMTD4 — the platform that can pausehttps://rajneeshkambhatla.com/essays/2026-06-09-d4-technology-platform/https://rajneeshkambhatla.com/essays/2026-06-09-d4-technology-platform/Technology and platform foundation becomes governable when the agent estate can route, throttle, fail over, and degrade without improvising under pressure.Mon, 08 Jun 2026 20:00:00 GMTD3 — what the model was allowed to learnhttps://rajneeshkambhatla.com/essays/2026-06-05-d3-data-knowledge/https://rajneeshkambhatla.com/essays/2026-06-05-d3-data-knowledge/Data and knowledge become governable when provenance, quality, residency, rights, and retrieval signals are visible at the point AI systems consume them.Thu, 04 Jun 2026 20:00:00 GMTD2 — what would change the decisionhttps://rajneeshkambhatla.com/essays/2026-06-02-d2-ai-to-value/https://rajneeshkambhatla.com/essays/2026-06-02-d2-ai-to-value/AI-to-Value becomes governable when every pilot names the evidence that would stop, narrow, pivot, or continue the work.Mon, 01 Jun 2026 20:00:00 GMTD1 — naming the AI leader, not the titlehttps://rajneeshkambhatla.com/essays/2026-05-29-d1-strategy-leadership/https://rajneeshkambhatla.com/essays/2026-05-29-d1-strategy-leadership/AI leadership becomes real when one named path can carry vision, board appetite, cost discipline, and competitive posture.Thu, 28 May 2026 20:00:00 GMTPillar 3 — enablement is not the soft pillarhttps://rajneeshkambhatla.com/essays/2026-05-26-pillar-3-enablement/https://rajneeshkambhatla.com/essays/2026-05-26-pillar-3-enablement/Enablement is the contestability layer: without meaningful HITL, reversibility and stop/pivot evidence collapse into theatre.Mon, 25 May 2026 20:00:00 GMTPillar 2 — governance is not a new theatrehttps://rajneeshkambhatla.com/essays/2026-05-22-pillar-2-responsible-agentic-ai/https://rajneeshkambhatla.com/essays/2026-05-22-pillar-2-responsible-agentic-ai/Responsible AI governance works when existing forums can pause, reverse, evidence, and notify before an agentic workflow outruns them.Thu, 21 May 2026 20:00:00 GMTStrategy as the negative-space disciplinehttps://rajneeshkambhatla.com/essays/2026-05-19-pillar-1-strategy/https://rajneeshkambhatla.com/essays/2026-05-19-pillar-1-strategy/AI strategy becomes real when the portfolio names what it will refuse, not only what it might govern.Mon, 18 May 2026 20:00:00 GMTAI governance after the maturity score: four decisions before the modelhttps://rajneeshkambhatla.com/essays/2026-05-12-meta-essay/https://rajneeshkambhatla.com/essays/2026-05-12-meta-essay/AI governance becomes real when it can change a decision before the decision becomes something the organisation has to explain afterwards.Mon, 11 May 2026 20:00:00 GMT