EDITOR’S NOTE: Today’s post (my response) is inspired by Stuart Winter-Tear’s LinkedIn post below.
This is how entire AI startup graveyards are made – one mere flick of the wrist from a tech behemoth.
I know several startups and consultancies building “RFP Agents.”
Months of work. Custom prompts. Pitch decks. VC calls.
Then Microsoft casually demos one at Build:
– Built in an afternoon.
– Writes in the firm’s voice and format.
– Pulls from SAP, DocuSign, Dynamics, etc.
– Runs compliance via multi-Agent orchestration.
– Posts final proposals into Teams – signed, sealed, delivered.
When platform primitives absorb your AI product thesis, you’re not disrupted, you’re deprecated.
This isn’t “feature vs product.” It’s distribution asymmetry, platform gravity, and the brutal efficiency of integration.
AI Startup New Playbook (Post-Copilot):
𝐁𝐮𝐢𝐥𝐝 𝐟𝐨𝐫 𝐭𝐡𝐞 𝐞𝐝𝐠𝐞 𝐜𝐚𝐬𝐞𝐬, 𝐧𝐨𝐭 𝐭𝐡𝐞 𝐝𝐞𝐦𝐨
Copilot nails the happy path. The value is in the complex, messy, regulated, or weird.
𝐎𝐰𝐧 𝐭𝐫𝐮𝐬𝐭, 𝐧𝐨𝐭 𝐣𝐮𝐬𝐭 𝐨𝐮𝐭𝐩𝐮𝐭
Enterprises don’t just want answers – they want traceability, compliance, and control.
𝐒𝐨𝐥𝐯𝐞 𝐫𝐞𝐚𝐥 𝐛𝐮𝐬𝐢𝐧𝐞𝐬𝐬 𝐛𝐨𝐭𝐭𝐥𝐞𝐧𝐞𝐜𝐤𝐬
If your AI doesn’t plug into a KPI, a budget, or a legal obligation – it’s not a tool. It’s a toy.
𝐆𝐨 𝐝𝐞𝐞𝐩, 𝐧𝐨𝐭 𝐰𝐢𝐝𝐞
Horizontal wrappers get absorbed. Vertical expertise – legal ops, finance, procurement, etc. – is where you survive.
𝐄𝐱𝐩𝐥𝐨𝐢𝐭 𝐭𝐡𝐞 𝐩𝐥𝐚𝐭𝐟𝐨𝐫𝐦 𝐛𝐥𝐢𝐧𝐝 𝐬𝐩𝐨𝐭𝐬
Liability. Governance. Market nuance. Tech giants won’t go near them. You should.
𝐃𝐨𝐧’𝐭 𝐛𝐮𝐢𝐥𝐝 𝐚 𝐰𝐫𝐚𝐩𝐩𝐞𝐫. 𝐁𝐮𝐢𝐥𝐝 𝐚 𝐰𝐨𝐫𝐤𝐟𝐥𝐨𝐰
Wrappers are disposable. Embedded workflows are sticky, defensible, and fundable.
Because when infrastructure becomes interface, the middle gets erased.
The only way out is up the value chain.
Remember: This is about the surface area hyperscalers deliberately avoid because it’s: too risky, too narrow, too slow to scale, not margin-friendly at their level.
That’s precisely where AI startups can step in with credibility, customisation, and courage.
Good luck.
MY RESPONSE
Multiple thoughts immediately come to mind regarding your comments, Stuart Winter-Tear.
To start, the 2007 Mendocino Project, which became DUET – https://bit.ly/3PZRyjs
Then we move on to Python in 2023 – https://bit.ly/3qRLTDh.
The Procurement Insights Archives contains a great deal more between 2007 and 2023, providing even greater detail and depth regarding the points raised in the Winter-Tear post.
That said, I understand, respect, and even admire Microsoft‘s efforts to continually look for ways to capitalize on its entrenched and loyal users.
However, as you point out, they are focusing on the surface, or as you call it, the wrapper requirements, without addressing the deeper and more significant challenges.
For example, taxonomy – https://bit.ly/44V2JmC
The above taxonomy post was inspired by the work of a longstanding solution provider, AdaptOne, and their essential work in addressing more than topsoil functionality. This speaks directly to your point of opportunity: human-led, agent-based problem solving versus myopic, tech-led, equation-based functionality.
30
Procurement Insights 
Tahj Bomar
May 26, 2025
I’d like to see the % of SMEs with a taxonomy.I bet most rely
piblogger
May 26, 2025
Here is a RAM 2025 Model 2 excerpt from the post regarding percentages:
Estimated Percentage
Large Enterprises: Likely 70–90% have a proper taxonomy, driven by regulatory needs (e.g., EU Taxonomy), ERP integration, and ProcureTech adoption in sectors like energy, utilities, and manufacturing. The EY report’s 88% compliance rate supports this for large European firms.
Mid-Sized Firms: Approximately 40–60% have structured taxonomies, often supported by solutions like AdaptOne or Tealbook, but many rely on partial or manual systems.
SMEs: Only 10–30% likely have proper taxonomies, as they often lack resources for advanced systems, relying instead on basic categorizations or no formal taxonomy.
Overall Estimate: Across all industries, considering the dominance of SMEs (which constitute ~90% of global businesses per World Bank data), I estimate 25–40% of companies have a proper taxonomy. This accounts for high adoption among large enterprises, moderate adoption among mid-sized firms, and low adoption among SMEs.