- 18
- Episodes
- Weekly
- Cadence
- 2026
- First episode
About Unlock the Pilot
Unlock the Pilot is about what happens after the AI demo works: the data nobody cleaned, the permissions nobody mapped, and the person who has to run it on a Tuesday morning. From Unlocked Consulting, a senior-led AI practice in Munich working across Europe and the US, drawing on systems in production and on published research. Pilots that stall, foundations, the EU AI Act in engineering terms, the real cost of an AI project. For the people at 20-to-500-person companies who have to make it work, and for the CIOs, CTOs and heads of AI who need evidence for their own board or stage.
- Publisher
- Unlocked Consulting
- Category
- technology
- Language
- en
- Explicit
- No
- First episode
- 1 Jul 2026
- Latest episode
- 16 Oct 2026
Latest episodes
18 episodes in the feed.
16 Oct 2026
The Visual Platform Tax
A content pipeline built on Make.com and Google Sheets took about nine months to get working. The happy path came together in a few weeks; the rest went to rate limits, silent node failures and retry logic expressed as more nodes. The rewrite in actual code took six weeks. This episode is about where that gap comes from, and why it shows up at company scale too. What you'll hear: Why first-build speed is the one measure every visual platform wins, and the wrong one How debugging in a canvas fails to amortize: knowledge doesn't carry from one connector to the next The bridge pattern of calling old flows as API endpoints, and why those endpoints never get revisited Why nine months of visual work was the specification that made a six-week rewrite possible A one-day test: have an engineer code your three ugliest flows before you commit a year Full article → https://unlockedconsulting.ai/blog/visual-platform-tax-no-code-automation-cost
10 Oct 2026
Europe cares most about AI sovereignty and depends on it least
A renewal questionnaire came back with a question nobody on the project had seen before: in which jurisdiction is the model executed, and who operates the hardware it runs on. Tracing the answer took the better part of a week, through one vendor to another company that actually ran the model, on hardware in a region picked eighteen months earlier for latency reasons and never written down. Part of the final answer was "we believe so." What you'll hear: Why sovereignty is now a text box on a vendor form, propagated downward through customer contracts Deloitte finds 32% of EMEA companies rely mainly on foreign AI stacks, against 11% in the Americas Why German boards treat this as an evidentiary problem, not a policy debate The three answers per system that make the questionnaire survivable, and which one always stalls What a 300-client Access migration teaches about separating what you depend on from who runs it Full article → https://unlockedconsulting.ai/blog/europe-ai-sovereignty-dependence-gap
10 Oct 2026
Your AI programme looks better from where you are sitting
Nobody in a room of two hundred people at an AI summit raised a hand when asked who had the authority to stop an AI system they oversee — not flag a concern, stop it. Every one of those organizations would tick yes on a survey about human oversight. The documentation exists, someone is named in it, and the answer in the room was still no. What you'll hear: Why directors and the wider workforce described the same companies 35 points apart in the same week How Deloitte's own respondents reported confidence in governing AI and broken operating models in consecutive answers Cloudera's split: 85% say the data strategy is defined, 18% say the data is fully governed Sixteen experienced developers measured a 19% slowdown after estimating a 20% speedup — having just done the work The comparison almost nobody runs: seats assigned against seats with activity in the last thirty days We end with one thing to do before the next round of AI spend: find out which instrument produced the case for the last one, then follow a single real request end to end with timestamps. Full article → https://unlockedconsulting.ai/blog/your-ai-programme-looks-better-from-where-you-are-sitting
10 Oct 2026
Everyone Measures AI Usage. 70% Can't Measure What It Returned
Anthropic surveyed 132 of its own engineers on Claude Code. Merged pull requests per day rose 67 percent, daily use climbed from 28 to 59 percent, and people reported productivity gains of 20 to 50 percent. Then someone looked at the delivery dashboard and nothing had moved. That gap is what this episode is about, and McKinsey's finding that only 30 percent of leaders can say where the freed-up time actually went says it is not one company's problem. What you'll hear: Why adoption and token spend are both costs, and neither answers the ROI question How "two hours saved per person" becomes dollars only if a budget line actually drops Why dollarizing a three-week calendar span invents labor hours nobody worked The METR study: a measured 19 percent slowdown against a self-reported 20 percent gain Two mechanisms that produce real money, and the different bridge each one needs Full article → https://unlockedconsulting.ai/blog/everyone-measures-ai-usage-70-can-t-measure-what-it-returned
10 Oct 2026
Vibe Coding Won't Kill Developers. It'll Kill the Middle
Cheap cameras didn't end professional photography. They hollowed out its middle: the small weddings, the headshots, the real estate listings. The same compression is working through software work right now, and the comfortable line — you were never paid to type, you were paid to think — only answers half of it. What you will hear: Why AI commoditizes a segment of software work rather than developers as a class The junior-to-mid tier lived on implementation against a known spec, which models now do cheaply Why "the middle learns to think better" confirms the argument instead of refuting it The capacity problem: the high end is narrow by definition and can't absorb everyone Where the photography comparison breaks, because software demand has grown for forty years We also get into what this means if you run a technical team: architecture, security, performance, and deep domain knowledge are exits from the middle, and senior-led work gets stronger as the middle thins. The closing exercise is simple. Look at the last five things your team shipped and count how many were implementation against a spec someone else wrote. Full article → https://unlockedconsulting.ai/blog/vibe-coding-wont-kill-developers-it-will-kill-the-middle
2 Oct 2026
Three studies measured agentic AI adoption. They got 23%, 59% and 20%.
Article: https://unlockedconsulting.ai/blog/three-studies-agentic-ai-adoption-different-numbers An internal slide carried one number, 59% of companies using agentic AI, and a bullet underneath saying we were behind. Two other studies from the same year said 23% and 20%. Nothing dramatic happened to the market between the fieldwork dates. Each of them counted a different thing. What the episode covers: What each figure actually counted: any use of an AI agent, at least one agentic use case in production, and autonomous systems running at scale against a metric someone instruments One internal system that was inside the count by the first definition and nowhere near it by the third Autonomy as a continuous variable sold as a binary, and three systems under the same label that break differently, cost differently and wake different people up The most useful comparison sitting inside one study: 59% with agents against 9% whose workflows run unattended The count that matters instead: how many workflows ran end to end last month with nobody approving a step, and someone acting on the result For operations leaders and founders at 20-to-500-person companies, and for the CIOs, CTOs and heads of AI who need evidence for their own board or stage. #AgenticAI #AIAdoption #EnterpriseAI
2 Oct 2026
The hours AI saved you have nowhere to go
Article: https://unlockedconsulting.ai/blog/hours-ai-saved-have-nowhere-to-go A manager built something with AI in an afternoon and handed the branch to engineering. It sat for months with nobody obstructing it, because review capacity had not changed and neither had the queue in front of it. AI accelerates the step in front of the person using it, and every step behind that person stays exactly as slow as it was. What the episode covers: AI spend up 110% in a single year while the share reporting integrated workflows across functions fell from 30% to 16%, the one integration figure in this year's research with a previous year to compare against 22,000 developers instrumented rather than surveyed: 34% more tasks finished, each one waiting 441% longer for review, and no change in how fast anything reached users Why tightening review is the wrong response when review was never the step that changed What happens to saved hours when nobody names a destination, and the 61% of 11,749 employees with limited or no guidance on what to do with them Three moves: decide where the hours go before buying, delete a handoff instead of speeding one up, and make each item cheaper to review For operations leaders and founders at 20-to-500-person companies, and for the CIOs, CTOs and heads of AI who need evidence for their own board or stage. #AIAdoption #DeveloperProductivity #EnterpriseAI
2 Oct 2026
Adoption Is a Latency Problem: The Zero-Backlog Policy
Article: https://unlockedconsulting.ai/blog/adoption-is-a-latency-problem-zero-backlog-policy The most common artifact of an enterprise AI rollout is a dead feedback channel. It opens in week one, collects suggestions for a month, and goes quiet by the third, because nothing anyone said ever came back as a release. Buy-in decays at the speed of your release cycle, and a suggestion sitting in a quarterly backlog proves the opposite of what the channel promised. What the episode covers: The zero-backlog policy as a promise to users: your suggestion ships within a day or two, or you hear the same day why it will not Putting the users on stage at company meetings instead of the project team, and why a testimonial and a pitch are distinguishable from the back of the room One business metric tracked from day one, finished videos per week, chosen because the business already watched it before the project existed Audit and traceability built into the first increment, so the governed path was never the slower one and nobody experienced it as a gate The taper: why zero backlog is a launch regime with a defined end, and what happens if you relax it before the ownership question is settled For operations leaders and founders at 20-to-500-person companies, and for the CIOs, CTOs and heads of AI who need evidence for their own board or stage. #AIAdoption #ChangeManagement #EnterpriseAI
2 Oct 2026
AI dragons: capable of everything, governed by nothing
Article: https://unlockedconsulting.ai/blog/ai-dragons-capable-of-everything-governed-by-nothing Seventy-five minutes produced a working agentic chatbot: chat interface, dashboard, tool access, and a model that could take a request and go do it. A week later the governance around it was still unfinished, and none of that week went into capability. The ratio is the shape of the work, and almost every budget for agentic AI has it backwards. What the episode covers: What the week went into: logging that outlives the session, attribution, explicit points where the system stops and waits, and boundaries on which systems, tables and operations it may reach Using the EU AI Act as a design brief rather than an obligation, on a system nowhere near the high-risk tier, and why that costs nothing while building Why every rule written turns out either loose enough to permit something unanticipated or strict enough to refuse work a person would have approved The distance between the constraint you described and the constraint the model inferred, and why it is always wider than it looks Why the taming behaves like a subscription rather than a project, and what the second line of an honest agentic budget says For operations leaders and founders at 20-to-500-person companies, and for the CIOs, CTOs and heads of AI who need evidence for their own board or stage. #AgenticAI #AIGovernance #EnterpriseAI
2 Oct 2026
You cannot fire your AI agents
Article: https://unlockedconsulting.ai/blog/ai-agent-identity-access-revocation A branch came up for review with about sixty commits, all carrying one engineer's git identity. He had looked at four or five of the decisions behind them. A new hire and a printer both have four things settled: who they are, what they can reach, what they did, and how you switch them off. The agent your team stood up last month has none of them. What the episode covers: The same four questions asked of a new hire, a printer and an agent, and where the agent stops having answers Why distinct identity works on API-native platforms and breaks on business systems priced per seat, where sharing a credential is what teams actually choose Attribution as three facts on every action: which agent, which run, and which human or event authorised that run The revocation test, and why "we would rotate the key" describes an action rather than an outcome Estonia's approved framework for verifiable agent identity from June 2026, not yet law in force, and whyretrofitting attribution onto a shared service account is expensive For operations leaders and founders at 20-to-500-person companies, and for the CIOs, CTOs and heads of AI who need evidence for their own board or stage. #AIAgents #AgenticAI #AIGovernance
2 Oct 2026
Financial Analysis Needs AI at the Edges. Never in the Math
Article: https://unlockedconsulting.ai/blog/financial-analysis-needs-ai-at-the-edges-never-in-the-mathFinancial analysis is probably the most deterministicworkload in the building. Every metric is an arithmeticexpression over named sums, and a formula allowed to beflexible has stopped being a formula. This is a pipelinewhere AI reads at the entrance, explains at the exit, andnever touches a number in between. The system described isa prototype; the pattern is the claim.What the episode covers: Three options for running analysis over the books, and why handing exported PDFs to a model and pointing one at the relational tables fail for the same reason one layer apart 33 metrics, 14 of them carrying thresholds, and the only two conditions in the system marked critical: trust money not covering deposits, and books that do not balance A second, cheaper model auditing the first one's parse,and the deterministic rule that decides when it runs so the quality check also controls the cost Prose written at the exit over evidence the code assembled, with a validator confirming every figure in the sentences traces back to a computed value What the system refuses to do: a missing report produces an empty metric with a reason attached rather than an estimate For operations leaders and founders at 20-to-500-personcompanies, and for the CIOs, CTOs and heads of AI who needevidence for their own board or stage.Two synthetic voices, generated with NotebookLM from anarticle written and reviewed by Helkyn Coello, founder ofUnlocked Consulting.#AIArchitecture #FinanceAutomation #EnterpriseAI
18 Sept 2026
AI Doesn't Think For You, It Thinks Like You
Article: https://unlockedconsulting.ai/blog/ai-doesnt-think-for-you-it-thinks-like-you (https://unlockedconsulting.ai/blog/ai-doesnt-think-for-you-it-thinks-like-you) AI output arrives looking like a neutral technicaldocument. It is a reflection of whoever wrote the prompt,put through a layer that makes the perspective behind itharder to see. The same internal tool, specced separatelyby a business manager and by an engineering team, producedtwo sound architectures with almost nothing in common.What the episode covers: Two specs for one requirement, one built around control and manual approval at every step, the other aroundautomated ingestion and ten times the current volume Why an AI-generated document gets read as technicallyneutral, and how that turns into a shield against engineering pushback Authority transfer: what is actually deciding "best" when someone says the AI recommended this approach A microservices architecture that was never asked whether the team could operate it, after five years of running monoliths Running the same problem through business, technical and security framings, and reading the trade-offs where the three outputs diverge For operations leaders and founders at 20-to-500-personcompanies, and for the CIOs, CTOs and heads of AI who needevidence for their own board or stage.Two synthetic voices, generated with NotebookLM from anarticle written and reviewed by Helkyn Coello, founder ofUnlocked Consulting.
18 Sept 2026
The Hidden Mental Model That Determines Whether AI Tools Help You
Article: https://unlockedconsulting.ai/blog/hidden-mental-model-ai-tools (https://unlockedconsulting.ai/blog/hidden-mental-model-ai-tools) A working knowledge base came together in fifty-fiveminutes during a company meeting. The tool was not thereason. The judgment about how the pieces had to fit wasalready there, and the AI executed against it at speed. Putthe same setup in front of someone without that picture andit runs just as fast, toward a structure nobody canevaluate.What the episode covers: What the model is concretely: where data originates, what talks to what, how something gets from a machine to where it runs The two things it keeps getting confused with, writingcode and knowing the business, and why neither is what is missing when AI output goes wrong Why "which AI tool should we use" is tempting: it is answerable, comparable, and every vendor is glad toanswer it What an evaluation across Cursor, Claude Code, Copilot and the rest actually settled, and what it could not The whiteboard test: hand your fastest and slowest person a marker and ask each to draw the system they work on For operations leaders and founders at 20-to-500-personcompanies, and for the CIOs, CTOs and heads of AI who needevidence for their own board or stage.Two synthetic voices, generated with NotebookLM from anarticle written and reviewed by Helkyn Coello, founder ofUnlocked Consulting.
18 Sept 2026
Is Your Knowledge Base Actually Thinking, or Just Retrieving?
Article: https://unlockedconsulting.ai/blog/knowledge-base-thinking-vs-retrieving (https://unlockedconsulting.ai/blog/knowledge-base-thinking-vs-retrieving) Embed-and-retrieve answers what a document says. Thequestions that justify the budget ask what a situationmeans, and those answers sit across documents that share novocabulary with each other or with the question. Similarityhas no way to represent the thing that connects them.What the episode covers: Why cosine similarity is a proxy for relevance, and the exact point where the proxy stops holding One real cross-source question, answered across a two-year-old support ticket, a training recording and an enhancement document What the second representation costs: entity extraction, entity resolution, and a schema someone who knows the domain has to define Why the cost difference between the two architectures is incremental while the capability difference is not The five-question test to run before approving any knowledge base design For operations leaders and founders at 20-to-500-personcompanies, and for the CIOs, CTOs and heads of AI who needevidence for their own board or stage.Two synthetic voices, generated with NotebookLM from anarticle written and reviewed by Helkyn Coello, founder ofUnlocked Consulting.
18 Sept 2026
AI in the Data, Code in the Process
Article: https://unlockedconsulting.ai/blog/ai-in-the-data-code-in-the-process (https://unlockedconsulting.ai/blog/ai-in-the-data-code-in-the-process) If you need the same answer twice, do not use AI.Non-determinism is the whole value at the data layer, wherea question asked three different ways should still find theright material. In process execution it is a defect, and aninventory calculation that varies by interpretation is notintelligence.What the episode covers: The difference between AI assisting during development and AI executing in production, and why their failure modes are not comparable Where interpretation earns its non-determinism: natural language queries, classification, synthesis across disconnected sources Where it becomes a bug: payment flows, validation rules, anything an auditor will ask about An orchestration platform specified by name for work that was entirely deterministic, and what came out of removing it 59% of organisations using agentic AI against 9% withautonomous workflows, and what sits in that gap For operations leaders and founders at 20-to-500-personcompanies, and for the CIOs, CTOs and heads of AI who needevidence for their own board or stage.Two synthetic voices, generated with NotebookLM from anarticle written and reviewed by Helkyn Coello, founder ofUnlocked Consulting.
18 Sept 2026
80% of Companies Are Already Out of Step With the AI Act and Don't Know It
Article: https://unlockedconsulting.ai/blog/ai-act-engineering-floor-most-companies-missing (https://unlockedconsulting.ai/blog/ai-act-engineering-floor-most-companies-missing) The EU AI Act does not ask high-risk system operators tostart logging and testing. It assumes they already do. TheEnterprise AI Maturity Index found that 20% oforganisations have implemented AI testing, auditing andrisk-assessment processes, which puts the other four fifthsbelow the floor the high-risk regime is written on top of. What the episode covers: What Articles 12, 14, 15, 18 and 19 take for granted before a single obligation applies The 20% figure and where it comes from: 4,500 executives and 2,000 employees across 19 countries and 12 industries Why 80% is an inference from that baseline and not a measured compliance number, and why the distinction matters Classification before panic: most internal tools sit in the minimal-risk tier and carry none of this The one-hour test: list every AI system touching employment, access, credit or essential services, then ask whether six months of logs could be produced for one decision For operations leaders and founders at 20-to-500-personcompanies, and for the CIOs, CTOs and heads of AI who needevidence for their own board or stage.Two synthetic voices, generated with NotebookLM from anarticle written and reviewed by Helkyn Coello, founder ofUnlocked Consulting.
18 Sept 2026
The On-Premise LLM Lottery
Article: https://unlockedconsulting.ai/blog/on-premise-llm-lottery (https://unlockedconsulting.ai/blog/on-premise-llm-lottery)Running an LLM on your own hardware is not a four-stepchecklist. It is a search across model size, quantizationand runtime, landing on hardware that procurement alreadyfixed before anyone asked. Twenty combinations were testedhere before one of them fit.What the episode covers: - Why the tutorials are accurate and still useless: each documents one point in the space, on one GPU that is never yours- The three axes that interact, and why a parameter count tells you nothing until it is paired with quantization and a runtime- Why twenty attempts is the size of the job rather than excessive diligence- The failure mode that arrives after the demo: a model that answers one user and falls over under concurrent requests- Why a finished search is hard to copy, and how to scope a small version of it before committing to on-premFor operations leaders and founders at 20-to-500-personcompanies, and for the CIOs, CTOs and heads of AI who needevidence for their own board or stage.Two synthetic voices, generated with NotebookLM from anarticle written and reviewed by Helkyn Coello, founder ofUnlocked Consulting.
1 Jul 2026
Can a Non-coder become a coder just with AI
Article: https://unlockedconsulting.ai/blog/can-non-coder-become-coder-with-ai (https://unlockedconsulting.ai/blog/can-non-coder-become-coder-with-ai) AI will write working code for someone who has never programmed. The problem arrives later, when that code has to change. Without a model of how the parts fit together, nobody can evolve it, debug it or defend it in production, and what looked like productivity turns out to be technical debt on a shorter fuse than usual. This episode is about the floor: the smallest mental model a non-developer needs before AI output can be judged instead of trusted. What the episode covers: Why "here are the tools" is the wrong way to onboard a non-developer Output that looks correct and breaks three layers down, and why that failure is invisible to the person who generated it Five concepts that decide whether AI multiplies a business team or generates debt for the engineering team Why the required model is much smaller than a CS degree, and where its lower bound actually sits What a technical leader should settle before handing these tools to non-developers For operations leaders, founders and technical decision-makers deciding whether to let non-developers build with AI.
Contact Unlock the Pilot
- Guest appearances
- Does not typically book guests
Based on episode analysis; this does not confirm that the show is currently accepting guests.
Host of Unlock the Pilot?
Claim your podcast to manage its listing and keep your show details accurate.
Pod Engine is an independent podcast discovery and analytics service and is not affiliated with or endorsed by this podcast. Artwork and show content belong to their owners. Full legal notice.
Explore this show
Podcast research with Pod Engine