“If you get a vendor to build a certain feature, you might have an exclusive period… but after that, everybody in the industry has that feature.”

— Bharat Bhushan, Partner, Tech & Data

Two recent observations got me fired up to write this post. First, at AIM Research’s recent CDO Vision event in San Francisco, data chiefs made it clear: “CDOs are prioritizing adding relevant capabilities over purchasing comprehensive platforms.” Likewise, a Slashdot story headlined “Enterprises Are Shunning Vendors in Favor of DIY Approach To AI” confirms that bleeding edge teams are embracing build over buy.

Second, I see a stubborn fear of technical debt keeping some execs tethered to old best practices. Million-dollar vendor bets are crumbling under GenAI’s relentless evolution, yet some leaders double down, burying innovation to save face. That’s not leadership… it’s denial.

Executives love preaching the virtues of buying GenAI. Let’s examine some of them:

a. Faster Implementation

Sure, you can spin up Microsoft CoPilot or a generic chatbot in weeks—but that’s so 2024. If you think competitive edge lies in off the shelf chatbots, you’re already walking toward obsolescence.

b. Reduced Development Overhead

Only if your GenAI lives in isolation. Modern agentic AI drinks from every enterprise system—CRM, ERP, you name it. Stitching a vendor’s tool into that mosaic still demands hardcore systems engineering. So much for “overhead saved”.

c. Scalability

Vendors love to tout infinite scale… until you’re shackled to their roadmap. Ever tried escaping Oracle Fusion Cloud ERP for SAP S/4HANA Cloud? Now picture that lock-in with GenAI, the most disruptive tech of our lives. Limited data security visibility, zero pricing leverage, and a front-row seat to vendor whims. Scalability? More like servitude.

Skeptical executives will raise objections like the ones below to building GenAI in-house. Let’s address them head-on:

a. “Higher Upfront Costs”

At $30 per user per month, a 1,000 seat CoPilot license runs $360K a year… forever. Meanwhile, citizen data scientists can stand up simple GenAI services on hyperscalers making API calls to low cost open source giants like llama3 70b or deepseek r1 distill qwen 32b.

b. “Requires Specialized Knowledge”

Yes, building custom AI solutions needs AI talent… but integrating a vendor’s tool into your sprawling enterprise stack isn’t a picnic either. Expertise is table stakes, not a dealbreaker. Plus, DIY keeps your talent sharp.

c. “Technical Debt”

If you think shadow IT doesn’t exist in your four walled palace, think again. Forcing vendor-only solutions fuels rogue projects. Instead, unleash citizen data scientists in a governed sandbox. Let them experiment, then scale what shines. That’s how you tame debt and spark breakthroughs.

d. “Risk of Failure”

DIY flops sting, but a multi-million-dollar SaaS disaster that misses ROI hurts way more. Worse still is the political fallout when a small scale, low cost prototype outperforms a bloated enterprise system.

This isn’t binary… it’s strategic. Here’s how to decide like a CAIO.

• Define the Problem First: If it’s repeatable and commoditized, buy. If it’s core to your value, build.
• Inventory Existing Software: Enterprise SaaS still plays silos like it’s 2015. Need bespoke glue code? Build, even if model context protocol (MCP) promises fixes someday. The build vs buy decision here depends on how much customization your stack demands.
• Assess Team Capacity: Do you have the right mix of data engineers, MLOps, and citizen scientists?
• Consider Long Term Costs: Tally maintenance, integration, talent, subscriptions, lock-in, and infra. Vendor deals look cheap until you’re trapped.
• Maintain Adaptability: GenAI evolves faster than your quarterly reviews. Even vendor tools should let you swap LLMs, vector stores, or pipelines on the fly. Stay nimble or get left behind.

Invest not just in technology, but in a culture of experimentation. Give your citizen data scientists a safe, governed sandbox. Harvest their ideas in an “innovation idea space,” then rapidly prototype the winners. That’s where you’ll find your exclusive GenAI edge, before your competitors catch up, or your key talent shifts to the enterprise that dares.

Stay tuned for my next article on turning sandbox successes into enterprise grade solutions. Thank you for listening in.