What 30 Years of Procurement History Tells Us About Glass’s G-Commerce
By Jon Hansen | Procurement Insights | January 2026
The Glass announcement for the “next generation of G-Commerce” is a well-crafted, timely, and aggressively positioned GovTech launch — one that rides the current AI-agent wave hard while targeting a genuinely underserved market: public-sector procurement. It’s smart marketing, credible early traction, and addresses real pain points in government buying.
However, it also repeats almost every structural flaw I’ve been documenting for 18 years in procurement technology — just dressed in public-sector compliance clothing.
Strengths: What They Got Right
Government-native constraints embedded. This is legitimately differentiated. Federal, state, and local procurement is drowning in FAR/DFARS, SAM registration, set-asides (AbilityOne, HUBZone, veteran-owned, women-owned, disadvantaged, Made in USA, green/sustainable), mandatory sources, and policy overlays. If their AI truly “reasons over” these in real time — eligibility checks, preferred-program routing, compliance validation — it removes massive manual burden.
Conversational guided buying. Natural-language input (“under $50k, 10-day delivery, HUBZone preferred”) translated into tailored, compliant recommendations is a meaningful leap over keyword and catalog searching. For non-expert buyers common in smaller agencies, this could be transformative.
Real traction numbers. $8M processed, 6M products delivered, $4.6M to small businesses, 18K+ users, 124 agencies — these are solid early-market metrics for a GovTech player. Partnerships with Santa Monica and South San Francisco add credibility.
Software category callout. Recognizing SaaS and licenses as a dedicated AI layer is smart. Government software procurement is notoriously painful (security, FedRAMP, pricing opacity); solving that could be their wedge.
Messaging discipline. “AI turns complexity into clarity, compliance into confidence” is crisp, executive-friendly, and values-aligned with transparency, accountability, and public trust. They avoid overpromising “revolutionary” and lean into mission support.
Weaknesses: The Pattern Repeating
Despite the polish, this is classic technology-first thinking with the same blind spots I’ve been calling out since 2007.
1. Readiness and capacity completely skipped.
Zero mention of assessing whether an agency is organizationally capable of absorbing this platform — governance maturity, change absorption, data discipline, knowledge transfer, stakeholder alignment, or pattern recognition of past failures. It’s all “the platform does X” without “is your agency ready to use X?”
This isn’t theoretical. In September 2007, I documented Virginia’s eVA initiative — a government procurement platform using Ariba SaaS. The technology worked. But eVA succeeded because Virginia led with process understanding, stakeholder collaboration, and departmental empowerment — not technology. The mandate came after buy-in, not before. Throughput went from less than 1% to over 80% in six years. Supplier engagement grew from 5,000 to over 14,000 vendors receiving orders.
Glass is launching “AI-first.” Virginia launched methodology-first. The technology was the same era’s equivalent. The outcomes won’t be.
2. New black box, same problem.
“AI guides them directly,” “real-time vendor eligibility,” and “instant document retrieval” sound great — but there’s no transparency into how the AI reasons, weights criteria, or handles conflicts (e.g., competing set-asides). Agencies can’t audit or challenge recommendations. It’s a faster, prettier black box.
The question isn’t whether the platform can guide decisions — it clearly can. The question is whether agencies can see why.
3. Vendor and API dependency trap.
8.5M products plus 5M punch-outs via APIs means heavy reliance on vendor data feeds — which can be incomplete, gamed, delayed, or manipulated. “Seamless” works until compliance breaks or a preferred vendor vanishes from the feed.
4. Same old illusion.
“Fundamentally changes how government buying happens” echoes every ERP, e-procurement, cloud, and digital transformation launch of the past three decades. Technology changes. The thinking — readiness last — doesn’t.
The 75-85% e-procurement failure rate I documented in 2007 hasn’t changed. Neither has the reason.
The Doom Loop Prediction
Glass is executing a strong GovTech play: credible metrics, smart positioning around compliance pain, timely AI branding, and real early traction. They could gain meaningful share in state, local, and federal niches — especially if the AI truly handles set-asides and policy logic at scale.
But it repeats the doom loop I’ve mapped across four technology eras:
New tech emerges → promises to fix procurement
Skips readiness and capacity assessment
Enforces new dependencies (APIs, AI black box, vendor feeds)
Same failure pattern, now faster
Blame shifts to the agency: “They weren’t ready”
Here’s what’s different this time: AI compresses time. It reduces friction. It increases throughput. Which means when readiness is missing, failure arrives faster and at greater scale.
Virginia avoided this loop by doing the opposite. They understood that government isn’t a single business — it’s Higher Education, K-12, Corrections, Public Safety, Transportation, Health, Social Services, and Construction, each with special needs, special rules, and special challenges. They avoided making eVA a “software project” and shifted the emphasis from cost justification to process understanding and refinement.
Glass has built impressive technology. Virginia built an impressive methodology — and then selected technology to support it.
The results speak for themselves.
The Bottom Line
Glass’s G-Commerce is a credible direction for public sector procurement if they can prove three things in real deployments:
Evidence over claims — auditability, sources, and decision trail for every AI recommendation
Governance over interface — decision rights, exception handling, and accountability before “guided buying” becomes the front door
Readiness over rollout speed — process, data, and role clarity before deployment
If those three are solid, “AI-first” becomes meaningful.
If not, it becomes the next version of the same story: same buying confusion, just faster.
The right question isn’t whether the platform works. It’s whether agencies are ready to absorb it. That’s not a technology problem. That’s a readiness problem. And it’s the one most implementations skip.
Virginia eVA Update 2026
For those interested in the Virginia eVA case study that demonstrates what methodology-first government procurement looks like — and its long-term validation:
Yes Virginia! There is more to e-procurement than software! (Part 1): https://bit.ly/3SUWxml
Yes Virginia! There is more to e-procurement than software! (Part 2): https://bit.ly/3qHEyq4
Revised Forrester Wave Report confirms what I have been saying since 2007… eVA is tops in public procurement: https://bit.ly/49S3XAQ
Virginia’s eVA has now operated for 25 years — through three technology generations (Ariba to Ivalua), processing $127 billion in purchases, serving over 1,100 government entities, and saving $25 million annually. The technology changed. The methodology didn’t. That’s the difference between a platform and a transformation.
The Technology Lesson
The fact that Virginia successfully migrated from Ariba to Ivalua in 2021 while maintaining continuity proves the Phase 0 thesis:
When you build methodology first, technology becomes interchangeable.
Glass is betting everything on AI-first technology. Virginia proved that technology is the replaceable part — methodology is the constant.
Procurement Insights
Glass G-Commerce: Strong Technology, Missing Methodology
Posted on January 13, 2026
0
What 30 Years of Procurement History Tells Us About Glass’s G-Commerce
By Jon Hansen | Procurement Insights | January 2026
The Glass announcement for the “next generation of G-Commerce” is a well-crafted, timely, and aggressively positioned GovTech launch — one that rides the current AI-agent wave hard while targeting a genuinely underserved market: public-sector procurement. It’s smart marketing, credible early traction, and addresses real pain points in government buying.
However, it also repeats almost every structural flaw I’ve been documenting for 18 years in procurement technology — just dressed in public-sector compliance clothing.
Strengths: What They Got Right
Government-native constraints embedded. This is legitimately differentiated. Federal, state, and local procurement is drowning in FAR/DFARS, SAM registration, set-asides (AbilityOne, HUBZone, veteran-owned, women-owned, disadvantaged, Made in USA, green/sustainable), mandatory sources, and policy overlays. If their AI truly “reasons over” these in real time — eligibility checks, preferred-program routing, compliance validation — it removes massive manual burden.
Conversational guided buying. Natural-language input (“under $50k, 10-day delivery, HUBZone preferred”) translated into tailored, compliant recommendations is a meaningful leap over keyword and catalog searching. For non-expert buyers common in smaller agencies, this could be transformative.
Real traction numbers. $8M processed, 6M products delivered, $4.6M to small businesses, 18K+ users, 124 agencies — these are solid early-market metrics for a GovTech player. Partnerships with Santa Monica and South San Francisco add credibility.
Software category callout. Recognizing SaaS and licenses as a dedicated AI layer is smart. Government software procurement is notoriously painful (security, FedRAMP, pricing opacity); solving that could be their wedge.
Messaging discipline. “AI turns complexity into clarity, compliance into confidence” is crisp, executive-friendly, and values-aligned with transparency, accountability, and public trust. They avoid overpromising “revolutionary” and lean into mission support.
Weaknesses: The Pattern Repeating
Despite the polish, this is classic technology-first thinking with the same blind spots I’ve been calling out since 2007.
1. Readiness and capacity completely skipped.
Zero mention of assessing whether an agency is organizationally capable of absorbing this platform — governance maturity, change absorption, data discipline, knowledge transfer, stakeholder alignment, or pattern recognition of past failures. It’s all “the platform does X” without “is your agency ready to use X?”
This isn’t theoretical. In September 2007, I documented Virginia’s eVA initiative — a government procurement platform using Ariba SaaS. The technology worked. But eVA succeeded because Virginia led with process understanding, stakeholder collaboration, and departmental empowerment — not technology. The mandate came after buy-in, not before. Throughput went from less than 1% to over 80% in six years. Supplier engagement grew from 5,000 to over 14,000 vendors receiving orders.
Glass is launching “AI-first.” Virginia launched methodology-first. The technology was the same era’s equivalent. The outcomes won’t be.
2. New black box, same problem.
“AI guides them directly,” “real-time vendor eligibility,” and “instant document retrieval” sound great — but there’s no transparency into how the AI reasons, weights criteria, or handles conflicts (e.g., competing set-asides). Agencies can’t audit or challenge recommendations. It’s a faster, prettier black box.
The question isn’t whether the platform can guide decisions — it clearly can. The question is whether agencies can see why.
3. Vendor and API dependency trap.
8.5M products plus 5M punch-outs via APIs means heavy reliance on vendor data feeds — which can be incomplete, gamed, delayed, or manipulated. “Seamless” works until compliance breaks or a preferred vendor vanishes from the feed.
4. Same old illusion.
“Fundamentally changes how government buying happens” echoes every ERP, e-procurement, cloud, and digital transformation launch of the past three decades. Technology changes. The thinking — readiness last — doesn’t.
The 75-85% e-procurement failure rate I documented in 2007 hasn’t changed. Neither has the reason.
The Doom Loop Prediction
Glass is executing a strong GovTech play: credible metrics, smart positioning around compliance pain, timely AI branding, and real early traction. They could gain meaningful share in state, local, and federal niches — especially if the AI truly handles set-asides and policy logic at scale.
But it repeats the doom loop I’ve mapped across four technology eras:
Here’s what’s different this time: AI compresses time. It reduces friction. It increases throughput. Which means when readiness is missing, failure arrives faster and at greater scale.
Virginia avoided this loop by doing the opposite. They understood that government isn’t a single business — it’s Higher Education, K-12, Corrections, Public Safety, Transportation, Health, Social Services, and Construction, each with special needs, special rules, and special challenges. They avoided making eVA a “software project” and shifted the emphasis from cost justification to process understanding and refinement.
Glass has built impressive technology. Virginia built an impressive methodology — and then selected technology to support it.
The results speak for themselves.
The Bottom Line
Glass’s G-Commerce is a credible direction for public sector procurement if they can prove three things in real deployments:
If those three are solid, “AI-first” becomes meaningful.
If not, it becomes the next version of the same story: same buying confusion, just faster.
The right question isn’t whether the platform works. It’s whether agencies are ready to absorb it. That’s not a technology problem. That’s a readiness problem. And it’s the one most implementations skip.
-30-
Share this:
Related