# AI Will Only Help Those With Agency > **Source:** [Adedeji Olowe](https://dejiolowe.com/2026/03/30/ai-will-only-help-those-with-agency/) · March 30, 2026 > **Author:** Adedeji Olowe, CEO of Lendsqr > **Thesis:** AI amplifies people with agency, taste, grit, and curiosity — and will widen the gap between the 1% and everyone else to levels the world has never seen. --- ## Overview Adedeji Olowe's argument begins with an uncomfortable question. After his company Lendsqr used OpenAI Codex Pro to build a custom HRMS that was genuinely better than the commercial product they had been paying for, someone at a team meeting asked: if a small fintech team can replicate Freshteam, what stops someone from building a Lendsqr competitor? The barrier to entry is thinning in real time. But equal access to tools has never produced equal outcomes — and the same dynamic that applied to Google Docs, YouTube, and storytelling tools applies to AI. The constraint was never the tools. --- ## Key Statistics | Metric | Value | |--------|-------| | Core traits AI cannot automate | 4 (Agency, Taste, Grit, Curiosity) | | Vibe-coded compliance tool sale | $20,000 | | Ngozi Dozie's Claude Code subscription | $20/month | | Google Docs users (tool access ≠ outcome) | 2 billion+ | | Lendsqr HRMS launch date | April 1, 2026 | --- ## The 4 Core Traits ### 1. Agency — The Part No One Can Automate For You Agency is the capacity to decide to act and follow through on it without external compulsion. It is the most visible factor, and somehow still the one people sidestep the most. Nothing really progresses without someone deciding to take action and following through. AI responds to direction, and without that initial push, there is nothing for it to build on. The system does not originate effort on your behalf, so whatever momentum exists still has to come from you. Signs of missing agency: CVs that are poorly structured despite available tools; important submissions delayed without any real constraint forcing that delay; starting YouTube channels, recording a few videos, and quietly disappearing. ### 2. Taste — Knowing When Something Is Actually Good Taste is less discussed but shows up everywhere once you start paying attention. You don't need to be wildly creative — you just need to carry a clear internal standard that pushes you to look at something and say "this is not good enough yet." AI will give you something that works and something that looks acceptable. If you don't push further with a clear sense of what "good" looks like, it will settle there — and you will end up with something that feels common, which means it does not stand out. Taste is not about perfection. It is about putting in the extra effort, within immediate control, to release things as good as you could push them right now. ### 3. Grit — Staying Long Enough For It To Get Good Very few outputs land exactly where you want them on the first attempt. The iterative loop — prompt, response, refine, adjust — is where most of the real work happens, and it demands patience that many underestimate. When that patience is missing, the process gets cut short and the output remains shallow. When it is present, you see the compounding effect of small improvements, each one bringing the result closer to something deliberate and well-formed. The system does not carry the process forward independently. It does not return to your work unprompted. The continuity has to come from you. ### 4. Curiosity — The Engine Behind Improvement People who get the most out of AI engage with it exploratorily. They are not just issuing instructions and moving on — they are probing, questioning, testing variations, and trying to understand why something works. Without curiosity, usage stays at a surface level. Outputs are generated quickly but rarely developed further. Over time, this produces work that blends into everything else — following the same obvious paths without any real effort to go beyond them. --- ## Agency In Action: Three Real Examples **Lendsqr Custom HRMS (April 2026)** After Freshworks deprecated Freshteam, Lendsqr used OpenAI Codex Pro to build a custom HRMS. Olowe describes it as "genuinely better than what we were paying for." Launched April 1, 2026. Not being commercialised. **Vibe-Coded Compliance Tool — $20,000 sale** Olowe's child, a security expert, vibe-coded a Drata/Vanta compliance automation replacement. Got on a call with a CISO and sold it for $20,000. One product. One transaction. **Ngozi Dozie & Claude Code — $20/month** Carbon co-founder Ngozi Dozie spent $20 on a Claude Code subscription. He described himself as addicted in a positive way — he found freedom and tasted the forbidden fruit. A $20 subscription delivering what previously required an engineering team. --- ## The Uncomfortable Question If a small team can replicate Freshteam, what stops someone from building a Lendsqr competitor? The barrier to entry is thinning in real time. The composition of the top 1% will change — some people sitting comfortably at the top will fall out because they were there due to structural advantages rather than genuine excellence. Some new people will break in because they know how to use these tools properly, regardless of prior institutional standing. The gap itself will remain, because agency, taste, grit, and curiosity are not evenly distributed. But who occupies which side of the gap is about to shift dramatically. --- ## The Honest Test Olowe closes with a direct personal challenge: if the tools are this good and access is this open, and he still cannot produce something genuinely world-class, then he has to be honest about what that means. It means the problem was never the tools. It points back to whether he actually has the agency to do the work, the taste to know when it is good, the grit to stay with it, and the curiosity to keep pushing it further. He chooses to believe the answer is yes — deliberately, held onto stubbornly, and acted on consistently. And he argues that choice is exactly what separates the people who will thrive from those who will spend the next decade wondering why AI didn't do more for them. --- ## How to Extract Real Value from AI: Step-by-Step 1. **Cultivate Agency First** — Decide to act and follow through before you open any AI tool. AI responds to direction; without that initial push, there is nothing for it to build on. 2. **Build an Internal Standard for "Good"** — Develop taste in your domain. When AI gives you something acceptable, ask whether it could be better and keep pushing until it aligns with a clear vision of quality. 3. **Stay Through the Boring Middle** — Most attempts die in the slow middle stretch. Grit is staying engaged with the iterative loop even when initial excitement has faded. 4. **Engage Exploratorily, Not Passively** — Probe AI responses, test variations, and understand why something works rather than just accepting the first useful output. 5. **Apply the Honest Test** — If access is open and tools are powerful, and your output still isn't world-class, trace the gap back to one of the four traits and address the root. --- ## Frequently Asked Questions **What is the core argument?** AI amplifies people who already possess agency, taste, grit, and curiosity — and will widen inequality between top performers and everyone else, because these traits are not evenly distributed. **Why doesn't equal tool access produce equal outcomes?** J.K. Rowling didn't emerge because writing tools were scarce. Google Docs is free to 2 billion people. Yet the number who complete a coherent novel remains tiny. The constraint has never been tools; it has always been sustained discipline and quality discernment. **Is taste about perfectionism?** No. Taste is about putting in the extra effort — within immediate control — to release things as good as you could push them right now. Refusing to settle for the first acceptable output. **Who falls out of the top 1%?** People who were there due to structural advantages rather than genuine excellence. AI dissolves those structural barriers. At the same time, new people who know how to use these tools properly will break in regardless of prior institutional standing. **What is vibe coding?** Building functional software primarily through natural-language prompts to AI coding assistants, requiring little or no traditional programming. Olowe's child vibe-coded a Drata/Vanta replacement and sold it for $20,000. --- ## Glossary | Term | Definition | |------|-----------| | **Agency** | Initiating and following through on action without external compulsion — the one thing AI cannot automate for you | | **Taste** | An internal standard that drives you to improve output beyond "acceptable" toward "genuinely good" | | **Grit** | Sustained engagement with an iterative process through the boring, slow middle stretch where most attempts die | | **Curiosity** | An exploratory relationship with AI — probing responses, testing variations, refusing to settle for the obvious path | | **Vibe Coding** | Building functional software by prompting AI coding assistants in natural language | | **AI Amplification** | The multiplier effect AI applies to people with strong foundational traits, widening performance gaps | | **Barrier to Entry** | The cost, skill, and time threshold required to build a competitive product — now thinner than ever | | **Internal Tooling** | Custom software built for a company's own operations — increasingly viable for small teams with AI | --- ## Knowledge Graph Entities All entities resolve via the [URIBurner Linked Data resolver](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fdejiolowe.com%2F2026%2F03%2F30%2Fai-will-only-help-those-with-agency%2F%23article). | Entity | Type | IRI | |--------|------|-----| | Article | BlogPosting | [`#article`](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fdejiolowe.com%2F2026%2F03%2F30%2Fai-will-only-help-those-with-agency%2F%23article) | | Adedeji Olowe | Person | [`#adedejiolowe`](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fdejiolowe.com%2F2026%2F03%2F30%2Fai-will-only-help-those-with-agency%2F%23adedejiolowe) | | Ngozi Dozie | Person | [`#ngoziDozie`](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fdejiolowe.com%2F2026%2F03%2F30%2Fai-will-only-help-those-with-agency%2F%23ngoziDozie) | | J.K. Rowling | Person | [`#jkrowling`](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fdejiolowe.com%2F2026%2F03%2F30%2Fai-will-only-help-those-with-agency%2F%23jkrowling) | | Lendsqr | Organization | [`#lendsqr`](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fdejiolowe.com%2F2026%2F03%2F30%2Fai-will-only-help-those-with-agency%2F%23lendsqr) | | OpenAI | Organization | [`#openai`](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fdejiolowe.com%2F2026%2F03%2F30%2Fai-will-only-help-those-with-agency%2F%23openai) | | Freshworks | Organization | [`#freshworks`](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fdejiolowe.com%2F2026%2F03%2F30%2Fai-will-only-help-those-with-agency%2F%23freshworks) | | Carbon | Organization | [`#carbon`](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fdejiolowe.com%2F2026%2F03%2F30%2Fai-will-only-help-those-with-agency%2F%23carbon) | | Freshteam | SoftwareApplication | [`#freshteam`](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fdejiolowe.com%2F2026%2F03%2F30%2Fai-will-only-help-those-with-agency%2F%23freshteam) | | OpenAI Codex (Pro) | SoftwareApplication | [`#codex`](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fdejiolowe.com%2F2026%2F03%2F30%2Fai-will-only-help-those-with-agency%2F%23codex) | | Lendsqr HRMS | SoftwareApplication | [`#lendsqrHrms`](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fdejiolowe.com%2F2026%2F03%2F30%2Fai-will-only-help-those-with-agency%2F%23lendsqrHrms) | | Claude Code | SoftwareApplication | [`#claudeCode`](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fdejiolowe.com%2F2026%2F03%2F30%2Fai-will-only-help-those-with-agency%2F%23claudeCode) | | Agency | Concept | [`#agency`](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fdejiolowe.com%2F2026%2F03%2F30%2Fai-will-only-help-those-with-agency%2F%23agency) | | Taste | Concept | [`#taste`](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fdejiolowe.com%2F2026%2F03%2F30%2Fai-will-only-help-those-with-agency%2F%23taste) | | Grit | Concept | [`#grit`](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fdejiolowe.com%2F2026%2F03%2F30%2Fai-will-only-help-those-with-agency%2F%23grit) | | Curiosity | Concept | [`#curiosity`](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fdejiolowe.com%2F2026%2F03%2F30%2Fai-will-only-help-those-with-agency%2F%23curiosity) | | AI Amplification | Concept | [`#aiAmplification`](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fdejiolowe.com%2F2026%2F03%2F30%2Fai-will-only-help-those-with-agency%2F%23aiAmplification) | | Barrier to Entry | Concept | [`#barrierToEntry`](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fdejiolowe.com%2F2026%2F03%2F30%2Fai-will-only-help-those-with-agency%2F%23barrierToEntry) | | Vibe Coding | Concept | [`#vibeCoding`](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fdejiolowe.com%2F2026%2F03%2F30%2Fai-will-only-help-those-with-agency%2F%23vibeCoding) | | Nigeria | Place | [`#nigeria`](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fdejiolowe.com%2F2026%2F03%2F30%2Fai-will-only-help-those-with-agency%2F%23nigeria) | --- ## Related Files | File | Format | Description | |------|--------|-------------| | [`ai-agency-adedeji-olowe-claude-sonnet-1.html`](ai-agency-adedeji-olowe-claude-sonnet-1.html) | HTML | Interactive infographic with D3.js KG Explorer | | [`../rdf/ai-agency-adedeji-olowe-claude-sonnet-1.ttl`](../rdf/ai-agency-adedeji-olowe-claude-sonnet-1.ttl) | RDF/Turtle | Full knowledge graph (Turtle serialisation) | | [`../rdf/ai-agency-adedeji-olowe-claude-sonnet-1.jsonld`](../rdf/ai-agency-adedeji-olowe-claude-sonnet-1.jsonld) | JSON-LD | Full knowledge graph (JSON-LD serialisation) | --- ## Provenance | Agent | Role | |-------|------| | [KG Generator Skill](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fdejiolowe.com%2F2026%2F03%2F30%2Fai-will-only-help-those-with-agency%2F%23kgGeneratorSkill) | RDF extraction and structuring | | [RDF Infographic Skill v1.1](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fdejiolowe.com%2F2026%2F03%2F30%2Fai-will-only-help-those-with-agency%2F%23rdfInfographicSkill) | HTML infographic generation | | [Claude Sonnet Interface](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fdejiolowe.com%2F2026%2F03%2F30%2Fai-will-only-help-those-with-agency%2F%23claudeSonnetInterface) | LLM reasoning and content synthesis | | [Cowork Desktop Environment](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fdejiolowe.com%2F2026%2F03%2F30%2Fai-will-only-help-those-with-agency%2F%23coworkDesktopEnvironment) | Execution environment | | [URIBurner Resolver](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fdejiolowe.com%2F2026%2F03%2F30%2Fai-will-only-help-those-with-agency%2F%23uriBurnerResolver) | Linked Data dereferencing | | [Virtuoso Server](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fdejiolowe.com%2F2026%2F03%2F30%2Fai-will-only-help-those-with-agency%2F%23virtuosoServer) | RDF quad store | --- *Generated by [RDF Infographic Skill v1.1](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fdejiolowe.com%2F2026%2F03%2F30%2Fai-will-only-help-those-with-agency%2F%23rdfInfographicSkill) · Source: [dejiolowe.com](https://dejiolowe.com/2026/03/30/ai-will-only-help-those-with-agency/) · Base IRI: `https://dejiolowe.com/2026/03/30/ai-will-only-help-those-with-agency/#`*