Every company, at some point, faces a fundamental build-vs-buy decision. Not everything can — or should — be built organically. Engineering capacity is finite, time-to-market matters, and some problems are already solved better by vendors who’ve invested years and millions into a specialized domain. The strategic call to buy rather than build is often straightforward. What’s far more complex — and frequently underestimated — is the buying process itself.
Procurement of enterprise software is messy, political, slow, and expensive. Even after leadership aligns on the decision to purchase, the journey from shortlisting vendors to full integration can span months or years, involve multiple stakeholders, and balloon in cost far beyond the initial contract value. Getting it right requires both a principled framework and the organizational will to execute it with speed and intentionality.
What a Real Enterprise Buy Looks Like
I’ll speak from direct experience. I was once tasked with leading a multi-million-dollar program to select, procure, and implement a market-leading personalization software platform for a large U.S. retailer. The scope wasn’t just picking a vendor — it included negotiating the RFP and Statement of Work, aligning internal delivery teams, designing the technical integration architecture, and rolling out the solution across three distribution centers, all within 12 months.
The integration itself was non-trivial. We connected the personalization platform with existing warehouse management applications using Apache Kafka and a microservices architecture — a setup that required deep coordination between the vendor’s team, our infrastructure engineers, and business stakeholders who had their own expectations about timelines and outcomes.
We delivered. But what I took away from that experience — and from several similar programs since — is that the outcome of a software buy is determined far upstream of the implementation. It’s determined by how you evaluate, select, and negotiate with vendors in the first place.
A Two-Principle Framework for Software Procurement
After years in this space, I’ve distilled enterprise software buying into two foundational priorities. Everything else — vendor demos, reference checks, security reviews, integration assessments — feeds into these two.
1. Bet on Innovators, Not Just Incumbents
The first and most important criterion is long-term innovation capacity. You are not just buying software for today’s use case. You are entering a relationship — often a multi-year, deeply integrated dependency — with a vendor. That vendor needs to still be relevant, financially healthy, and actively developing their product three, five, and ten years from now.
This means looking beyond feature checklists. A product that ticks every requirement box today but is built by a company with stagnating R&D investment, shrinking customer growth, or mounting technical debt is a liability in disguise. Conversely, a vendor who is slightly ahead of your current needs but has a clear innovation roadmap, strong financial backing, and a growing ecosystem is likely a safer long-term bet.
Shortlist vendors who are market leaders not by past reputation, but by demonstrated trajectory. Ask hard questions: What percentage of revenue goes to R&D? What is the product roadmap for the next 18 months? How do they incorporate customer feedback into their release cycle? Who are their design partners?
2. Project the True Total Cost — Then Negotiate Accordingly
Software procurement teams consistently underestimate the full cost of a purchase. The license fee or SaaS subscription is just the entry point. Over a multi-year lifecycle, enterprises typically spend significantly more on add-ons, customizations, professional services, training, support tiers, and integration overhead than they do on the base product.
Before signing anything, build a multi-year total cost of ownership model. Include: the initial license or subscription fee, implementation and integration costs (often 0.5x to 2x the license cost), annual maintenance and support, anticipated feature add-ons or usage-based expansion, and internal engineering time for ongoing maintenance.
Once you have that number, pressure-test whether the investment is ROI-positive within a reasonable horizon — typically three to five years. If it is, you have a negotiating foundation. If it isn’t, you either need to revisit the vendor, renegotiate terms, or escalate to leadership for a strategic investment justification. Some software buys are not ROI-positive in the traditional sense — they’re infrastructure or compliance investments that enable the business to operate. That’s a legitimate category, but it needs to be explicitly framed as such, not buried in optimistic assumptions.
Why These Principles Matter More Than Ever for AI Adoption
The same framework applies directly — and urgently — to the wave of AI software investment that CTOs and technology leaders are navigating right now.
AI adoption is no longer optional for most enterprises. The pace of innovation in large language models, coding assistants, AI-powered analytics, and autonomous workflow tools has reached a point where delayed investment creates compounding competitive disadvantage. Companies that are not experimenting with and deploying AI tools today are falling behind — not in a distant, theoretical sense, but in measurable productivity, speed, and capability gaps relative to early adopters.
This urgency creates a risk: organizations feeling pressure to move fast often skip the evaluation rigor that good software procurement requires. They sign multi-year contracts with vendors who look impressive in a demo but lack the staying power to be meaningful partners in two years. Or they invest heavily in tools before validating that the use case delivers real returns.
The two-principle framework cuts through this noise.
Applying Principle 1 to AI: Bet on Infrastructure-Level Players
In the current AI landscape, the vendors best positioned to remain relevant and continue innovating are those with the compute infrastructure, talent pipelines, and financial depth to compete at the frontier. For most enterprises, this means anchoring their AI stack to partners like Google (Vertex AI, Gemini), Microsoft (Azure OpenAI, Copilot ecosystem), and Amazon (AWS Bedrock, Q) — not because they are always the most cutting-edge on any individual benchmark, but because they are the least likely to disappear, pivot, or lose access to the compute resources required to run frontier models.
This doesn’t mean ignoring specialized vendors. For specific use cases — code generation, document intelligence, customer service automation, supply chain optimization — purpose-built AI vendors may significantly outperform the hyperscalers on the narrow problem you’re solving. But the anchor of your AI investment portfolio should be with players whose survival and continued innovation is close to certain.
Applying Principle 2 to AI: Start Small, Validate Fast, Scale Intentionally
AI software costs are often more opaque than traditional enterprise software. Token-based pricing, model access tiers, fine-tuning costs, inference infrastructure, and prompt engineering overhead are all real cost vectors that don’t show up in headline pricing. Before committing a significant budget, organizations should run structured proof-of-concepts on their highest-value use cases.
The POC approach serves two purposes: it validates that the AI tool actually delivers measurable improvement on your specific problem (not just impressive outputs in a controlled demo), and it surfaces the true cost-per-outcome before you’re locked into a large contract. A modest initial investment — enough to run a credible 60 to 90 day experiment across a real workflow — gives you the data to make a much more defensible scaling decision.
If the POC demonstrates clear productivity or revenue impact, you have the internal justification to expand investment and negotiate from a position of informed demand rather than speculative enthusiasm. If it doesn’t, you’ve saved the organization from a costly multi-year commitment to a tool that didn’t fit your context.
The Discipline to Move Fast Without Moving Blind
There’s a real tension in enterprise AI adoption right now. The fear of being left behind is driving urgency, but urgency without discipline leads to expensive mistakes. The organizations that will extract the most value from AI investment over the next decade are not the ones who signed the biggest contracts in the recent years — they’re the ones who established a principled evaluation process, validated use cases rigorously, and built vendor relationships with partners capable of sustained innovation.
The principles that govern good software procurement haven’t changed. What’s changed is the cost of getting it wrong. In a world where AI is becoming foundational infrastructure, a poor vendor selection or an unvalidated use case investment isn’t just a sunk cost — it’s an anchor on your organization’s ability to compete.
Buy deliberately. Evaluate rigorously. Start with proven innovators. Project the real cost. And move as fast as your discipline allows.
