About
I work at the intersection of agentic AI, composable architecture, and accessibility — currently as VP of Agentic Systems at Orium (opens in new tab) and Program Lead for the Agent Ecosystem at MACH Alliance (opens in new tab). This page tracks my writing across industry sites, partner blogs, and my LinkedIn profile (opens in new tab).
Writing
August 2026
WebMCP and Agentic Commerce (opens in new tab)
Published on MACH Alliance, August 12, 2026
WebMCP is a real technical advance — it lets a site declare the actions an agent can take instead of leaving the agent to scrape and guess at a page — but declaring an action isn't the same as being ready for agents to use it. Measured against three tests an emerging standard has to pass — adoption, enforceability, and reversibility — it currently comes up short on all three, which makes it something to watch rather than something to build a strategy on. The durable move is the same either way: composable, API-first foundations that outlast whichever protocol wins, with cautious low-risk experimentation for organizations already modernized enough to observe how agents actually behave.
Would You Trust a B2B Agent to Negotiate With Your Budget? (opens in new tab)
Published on Orium, August 5, 2026
Building a working prototype of a B2B negotiating agent revealed that safe autonomy takes more than functional code — it demands realistic decision-making frameworks, hard guardrails on sensitive information, and testing across many scenarios. Trustworthy negotiating agents need genuine strategic choices, enforced message filtering to prevent manipulation, and human oversight, especially around payment authorization.
Accountability Doesn't Come From Counting Vendors (opens in new tab)
Published on MACH Alliance, August 5, 2026
The instinct to consolidate vendors in the name of accountability mistakes a smaller supplier list for a clearer chain of responsibility — one throat to choke is not the same as knowing what broke. Accountability is an operational discipline instead: define the standard, measure against it, monitor the thresholds, trace failures back to their source, and put a named person on each one. Agentic systems raise the stakes by taking the cast from dozens of vendors to thousands of actors, but they don't change where responsibility ultimately sits, which is inside your own leadership rather than in anyone's contract.
July 2026
Agent-Driven Delivery Is Fast, Until Your Environment Slows It Down (opens in new tab)
Published on Orium, July 28, 2026
While AI coding agents can rapidly produce working code, most enterprises fail to realize the speed gains because bottlenecks exist throughout the rest of the delivery pipeline. The real question isn't which AI coding tool to pick, but whether an organization's structures and technical infrastructure can actually accommodate the velocity agents provide.
Lakshmi Showed Up, and So Did the Pull Requests
July 11, 2026
keyboard-a11y-tester gains a second persona, Lakshmi, who computes what a screen reader would announce from a page's live accessible tree rather than driving an actual screen reader. The pilot also earned OpenSSF Best Practices and Scorecard badges and its first outside pull requests within days of launch.
Revisiting AI Agents for Web Accessibility: What a keyboard-only testing agent proved, disproved, and deliberately refused to fake (opens in new tab)
Published on LinkedIn, July 8, 2026
A 2025 proposal imagined AI personas simulating disabled users' lived experience; this working pilot narrows that to one persona, a keyboard-only user, and finds the original framing needed correcting rather than confirming. A deterministic runner drives a real browser using only the keyboard, producing evidence-linked findings against WCAG success criteria like focus indicators, keyboard traps, skip links, and tab order, while the AI model is reserved for judgment calls a scanner can't make. The tool is explicit about what it refuses to fake or assert, positioning itself as a first-pass instrument that complements professional accessibility testers with disabilities rather than replacing them.
Before You Give an Agent the Keys (opens in new tab)
Published on Orium, July 6, 2026
Agentic AI security isn't a problem to be solved but a balance to be struck: anything that can act can act wrongly, so organizations need to calibrate each agent's autonomy to the damage it could cause. The piece identifies six areas where that tradeoff concentrates — credentials, data flow inspection, tool access, host isolation, review processes, and data retention — and argues that instrumentation and human accountability have to back up the technical controls.
Trusting Agent-Driven Delivery in Production (opens in new tab)
Published on Orium, July 3, 2026
Agent-driven software delivery shows real promise but keeps stalling in practice, and the fix isn't a better model. The path to trustworthy, production-ready output is removing friction from the delivery lifecycle: documenting conventions, building sandboxes, automating checks, and giving teams clear visibility into what agents changed and why.
MACH Alliance Maturity Assessment, 2026 Edition (opens in new tab)
Published on MACH Alliance, July 3, 2026
A discovery-stage orientation for enterprise technology leaders, reframing the MACH Alliance's principles as properties of a system — open, composable, connected — rather than a product checklist, and arguing that architecture has become the common denominator behind cost discipline, delivery velocity, and whether AI actually returns anything. The assessment spans six pillars, from strategy and governance through people, technology, and business intelligence, with each one extended to account for what agents need in order to participate: documented sources, real-time integration, standardized protocols, and programmatic rather than UI-only data interfaces. The 2026 revision updates the 2023 working group's framework for the agent ecosystem, on the premise that without sound foundations agents have nothing reliable to work with.
June 2026
A Better Search Bar, or a Relationship? (opens in new tab)
Published on Orium, June 29, 2026
Shopping agents are being built almost entirely around product discovery, but that undersells what they could be. Extending agents past checkout into onboarding, problem resolution, and ongoing personalized recommendations is what turns a one-time transaction into a lasting customer relationship.
May 2026
The First Wave of Agentic AI Is Already in Production (opens in new tab)
Published on MACH Alliance, May 7, 2026
Five enterprises have already moved AI agents into production with measurable results, and the pattern cuts against conventional wisdom: the biggest wins came from narrowly targeting specific operational bottlenecks, not from obvious use cases like chatbots. Composable, API-first infrastructure and tightly scoped starting points were common to every success story.
April 2026
Building the Agent Ecosystem: How We're Learning Together (opens in new tab)
Published on MACH Alliance, April 9, 2026
MACH Alliance's Agent Ecosystem initiative brings enterprises, practitioners, and technology partners together to build enterprise AI solutions collaboratively rather than in isolation. The program is organized around four themes — engagement, awareness, solution demonstrations, and trust-signal creation — with separate participation paths for business leaders and technical architects.
March 2026
Competitive Compression: The New Race for Execution Capacity (opens in new tab)
Published on Orium, March 13, 2026
Agentic AI is compressing the gap between intent and action, and that speed is becoming the new competitive battleground. Organizations with strong internal structure can turn that compression into an advantage, while those with fragmented processes will find their weak points exposed faster than ever.
What Agentic Systems Are Forcing Organizations to Confront (opens in new tab)
Published on Orium, March 4, 2026
As agentic AI moves from pilot to mainstream, it's surfacing five hard operating questions at once: competitive acceleration, operational clarity, workforce evolution, digital delivery, and shifting customer expectations. Adopting the technology isn't the hard part — the real work is fixing the structural issues it exposes.
December 2025
Move Beyond Digital Transformation with Agentic Thinking (opens in new tab)
Published on Orium, December 16, 2025
Digital transformation has become table stakes rather than an advantage, so the real differentiator now is agentic transformation: building systems, human and AI alike, that act autonomously and adaptively instead of just digitizing existing processes. That means designing for context-awareness and proactive action, not just faster versions of the old workflow.
November 2025
The Fluent Builder MCP Server: Accelerating Developer Productivity with Context-Aware Tools (opens in new tab)
Published on Fluent Commerce, November 25, 2025
Fluent Commerce and Orium built the Fluent Builder MCP Server, wiring AI-assisted development directly into the Fluent CLI so developers can generate GraphQL queries and modify rules and workflows faster. It's a concrete example of the Model Context Protocol — a year old at the time of writing — connecting agents to real developer tooling rather than staying a demo.
September 2025
Agentic Payments Explained: ACP, AP2, and x402 (opens in new tab)
Published on Orium, September 29, 2025
Three new protocols — ACP from Stripe and OpenAI, AP2 from Google, and x402 from Coinbase — are emerging to let AI agents transact autonomously. Rather than competing head-on, each sits at a different layer of the agentic commerce stack: checkout integration, trust and authorization, and programmable payments, respectively.
July 2025
1,000 Pilots Don't Make a Strategy (opens in new tab)
Published on Orium, July 29, 2025
Many organizations end up with a portfolio of successful agentic AI pilots that never add up to enterprise-wide impact. Getting past that plateau takes pairing bottom-up experimentation with top-down strategy, investing in shared platform infrastructure, and building an adaptive roadmap instead of testing isolated use cases forever.
Piloting Your Agentic Future (opens in new tab)
Published on Contentstack, July 18, 2025
Successful AI agent pilots in content workflows start small and keep humans actively in the loop rather than trying to automate everything at once. The most effective approach treats agents as collaborative partners on high-friction tasks, building trust incrementally instead of positioning AI as a wholesale replacement.
June 2025
Adaptive by Design: Building Commerce Systems That Respond in Real Time (opens in new tab)
Published on Composable.com, circa June 2025
Adaptive commerce systems use real-time data, AI, and composable architecture to keep adjusting offers, pricing, and fulfillment as customer behavior and market conditions shift, rather than relying on static, one-time personalization. Getting there requires tightly integrating data and experience layers and organizing teams for continuous experimentation, not a one-off redesign.
May 2025
The Future of Content Is Collaborative (opens in new tab)
Published on Composable.com, circa May 2025
Agentive AI is reshaping content workflows into a partnership rather than a straight automation play, accelerating planning, creation, editing, and distribution while people retain oversight and creative direction. Organizations applying this across the whole content supply chain — not just one stage of it — are seeing meaningfully better ROI.
Ecommerce: Humans + AI Agents at Work (opens in new tab)
Published on Fluent Commerce, May 1, 2025
AI agents in retail operations are evolving from narrow automation tools into genuine collaborative teammates. Composable order management systems make this practical by letting agents handle routine, well-defined work while people focus on judgment calls and refinement.
April 2025
Inclusive Teams, Intelligent Systems: Why DE&I is the Secret to High-Performing Human+Agent Collaboration (opens in new tab)
Published on LinkedIn, April 15, 2025
Diversity, equity, and inclusion aren't a side concern for human-AI teams — they're a precondition for the teams working well at all. Research cited here shows human+agent teams produce measurably better output when the surrounding culture is psychologically safe and people feel free to question what the AI recommends.
March 2025
What Stripe's $1.4T Says About Commerce's Future (opens in new tab)
Published on Composable.com, circa March 2025
Stripe's $1.4 trillion in 2024 transaction volume points to five shifts already reshaping commerce: AI as a direct revenue driver, AI agents making autonomous purchases, stablecoins enabling global payments, fraud sophisticated enough to demand networked defenses, and business models that now need continuous reinvention rather than periodic refreshes.
February 2025
Why 70% of Enterprises Are Choosing Composable DXP (opens in new tab)
Published on Composable.com, circa February 2025
Enterprises are moving away from monolithic DXP suites like Adobe and Sitecore toward composable, API-first, cloud-native alternatives that cost less to run and flex more easily. Platforms like Builder.io, Contentful, and Contentstack are leading that shift by letting organizations assemble best-in-class tools instead of committing to one vendor's whole stack.
Leveraging Agentive AI for Web Accessibility: A New Frontier (opens in new tab)
Published on LinkedIn, February 1, 2025
Static accessibility checklists miss a lot of what actually breaks for disabled users, so this piece proposes AI agents that simulate specific disability experiences — limited mobility, color blindness, autism — and report back in narrative form. The goal is testing that reflects real, lived friction rather than a compliance box-check.
January 2025
Accessibility's Next Chapter: How Coding Assistants, Adaptive Overlays, and Agentive AI Personas Will Reshape Inclusivity (opens in new tab)
Published on LinkedIn, January 7, 2025
Tools like GitHub Copilot show real promise for pushing web accessibility past current standards, but code generation alone isn't the whole answer. Real inclusion still depends on balancing automation with human insight and testing against actual user experience, not just automated conformance checks.
November 2024
Creating a Data Fabric for Composable Commerce (opens in new tab)
Published on Composable.com, circa November 2024
A data fabric ties together the fragmented data that piles up across composable commerce stacks, enabling real-time sync instead of batch reconciliation. Built on services like Pub/Sub, BigQuery, and Vertex AI, it gives brands centralized governance and AI-driven personalization without giving up the flexibility of best-of-breed tools.
July 2024
Empowering Healthcare Businesses with Composable Commerce (opens in new tab)
Published on Composable.com, circa July 2024
Healthcare organizations can adopt composable commerce to build modern, engaging digital experiences while keeping patient data compliant with HIPAA — by isolating sensitive data in secure backend systems and assembling specialized, best-of-breed tools around it. That lets providers compete with consumer B2C brands on experience without loosening regulatory or security requirements.
January 2024
Master Composable Commerce with Free Starter Tools (opens in new tab)
Published on Composable.com, circa January 2024
Headless and composable commerce is more approachable than it sounds, thanks to free trials from vendors like commercetools, Algolia, Vercel, Stripe, and Contentstack. Composable UI in particular offers developers an open-source, no-cost way into the ecosystem without a big upfront commitment.
Unified Omnichannel Checkout: Retail's Competitive Edge (opens in new tab)
Published on Composable.com, circa January 2024
A unified checkout across physical stores, digital channels, and emerging platforms reduces cart abandonment and gives retailers a real competitive edge. Combining seamless purchasing paths with shared customer data lets retailers personalize the experience in ways that build conversions and long-term loyalty, not just a one-time sale.