There’s no shortage of talk about what AI agents can do these days, or what they are supposed to do. But examples of actual Agentic AI usage beyond testbeds and labs remain scarce across businesses. To share a few concrete examples, let’s look at autonomous sourcing in procurement, where agentic AI fits now, how to run it safely, and how to move from pilots to scaled impact without spooking your teams or regulators. The following insights are a summary of a panel discussion I recently moderated for procurement leaders.
Start Small and Set the Guardrails Early
Treat an AI agent like a new analyst on day one. You wouldn’t assign them to the make-or-break category. You would give them a low-risk, fragmented tail-spend event with clear rules, watch how they work, and iterate. The same applies here. Begin in categories where qualified suppliers already exist and award logic is straightforward, such as price plus one or two additional factors like emissions or quality. Codify those factors as policy: bonus/malus for incumbency, minimum performance thresholds, approved negotiation levers, and walk-away conditions. Then lock the autonomy mode to “recommend-only.” The agent drafts RFP questions, suggests suppliers, runs the bid, even proposes counter-offers. You review, override if needed, and approve.
Two non-negotiables: traceability and auditability. Every action the agent takes should be explainable, detailing the data it used, the alternatives it considered, and why it chose a certain recommendation. That audit trail is your safety net for internal assurance and evolving regulations (including “human-in-command” expectations). Finally, build a rhythm of retrospectives. After each cycle, ask: where did the logic fall short? Did we overestimate the cost? Did an incumbent bonus distort outcomes? Tune the rules, expand the data the agent can access, and widen the scope only when you see consistent results.
Thank you for reading The AI MEMO. Feel free to share it with your network.
NEW ONLINE COURSE — How AI Agents Augment Human Skills in the Workplace
Discover how to integrate agentic AI effectively into your work processes to boost productivity and decision-making. Learn how to select appropriate AI tools, delegate tasks, and verify outputs for high-quality results to prevent AI workslop. This course covers understanding limitations, biases, and ethical considerations in AI-generated content, equipping you to align outputs with your goals. Elevate your professional standards and prepare for the future of work by mastering the balance of AI assistance in your daily tasks.
The fastest way to earn buy-in is to show cycle time compression at the front end of source-to-contract. Agents excel at scope shaping, supplier discovery, templating, and parallel communications—areas that often lead to multi-day handoffs. Time those steps and publish the before/after. Next, track touchless spend rate across the event: automated comparisons, scenario modeling, award recommendations, and award-to-contract creation. Each percentage point you move from manual to touchless increases your team's capacity.
Don’t wait to have the perfect data. Pilot with what you have to reveal what you need: cleaner supplier master attributes, more explicit PO descriptions, standardized RFP templates. Good news: modern AI helps normalize messy inputs, so you don’t need a six-month cleanse to start. Do expand the agent’s field of view beyond prices—bring in supplier performance, risk, sustainability, and even prior negotiation transcripts. Based on historical purchase trends, an agent can infer viable counteroffers and the timing of concessions. With service contracts, it can learn reasonable variation year-over-year and flag outliers. Better award logic grounded in more signals convinces finance and risk partners to lean in. And remember: the goal isn’t to “beat” the counterparty; it’s to land a favorable, sustainable business outcome. Your guardrails should encode that ethos.
Scale Through People with Skills, Roadmap, and Organizational Buy-In
Technology is the easy part. Leaders should spend even more time on their people, who get relief when tactical load drops and time opens for supplier strategy, risk planning, and value creation. Set that narrative early. Position agents as teammates who take on the repetitive work so your experts can focus on higher-value work. Upskill on two fronts: conversational interaction with AI (clear instructions, structured prompts, review discipline) and data literacy (interpreting agent outputs, spotting bias, reading trendlines). The learning curve is far gentler than a legacy IT rollout, but it’s still a muscle.
Build your roadmap in three phases:
Phase 1: fix the intake (no more free-form emails), centralize the data feed, and run a tail-spend pilot with recommend-only autonomy.
Phase 2: expand to non-contracted spend in a defined category; introduce award scenario rules and automated award-to-contract.
Phase 3: layer in predictive elements like tariffs, risk signals, and performance scores to sharpen supplier selection and negotiation levers.
Each phase should end with a visible win you can share within your organization. Use those proof points to show adjacent teams what’s possible, then transpose the pattern to new categories (e.g., transport or direct materials). Throughout, keep compliance top of mind: document human-in-command checkpoints, approvals, and logs. Most importantly, secure executive sponsorship early. When leaders commit to the vision, the rest falls into place: budget, change management, and support.
Conclusion
Start small with agentic AI in low-risk events, with tight guardrails and full traceability. Prove value with cycle-time compression and touchless spend gains, fueled by “good enough” data and smarter negotiation signals. Then scale through people: upskill, phase your roadmap, and secure buy-in. Do this, and procurement stops being a back-office cost gate and becomes a strategic growth partner—faster, more accurate, and frankly, more fun. Your next step: pick one category, one event, one KPI. Pilot it. Learn fast. Expand. I’m here to help you make it real.
Need help setting up your AI projects and avoid sunk cost? Get in touch!
Thank you to KonnectHouse for hosting the event and to Zycus for sponsoring it.
Join my bi-weekly live stream and podcast for leaders and hands-on practitioners. Each episode features a different guest who shares their AI journey and actionable insights. Learn from your peers how you can lead artificial intelligence, agentic AI, generative AI & automation in business with confidence.
December 16 - Jon Reed (Industry Analyst and Co-Founder of diginomica) and I will wrap up 2025 with our own Agentic AI recap and a 2026 outlook. [More details to follow on my LinkedIn profile…]
Watch the latest episodes or listen to the podcast
