Why We Built 105 Free AI Agents for Infor
We built 105 AI agents for the Infor ecosystem and made every one of them free, because the fastest way to fix an underserved market is to remove the price of entry. Infor's platforms, SyteLine, LN, M3, Baan, ServiceMax, and the ION integration fabric, run tens of thousands of manufacturers, yet the modern AI tooling wave has almost entirely passed the ecosystem by: the consultancies sell hours, the platform vendors sell cloud subscriptions, and the practitioners doing the actual work get generic chatbots that hallucinate IDO names. This post explains the reasoning behind the giveaway, how Netray makes money anyway, and why we planted our flag on on-premise-first when the rest of the industry bet on cloud APIs.
The Infor Ecosystem's AI Tooling Gap
If you develop on Salesforce or SAP, you have an embarrassment of AI tooling: vendor copilots, startup ecosystems, fine-tuned models, community prompt libraries. If you develop on SyteLine or LN, you have general-purpose chatbots that invent IDO properties, suggest editing Baan sources that no longer exist, and cheerfully recommend direct table writes that would corrupt your ERP. The ecosystem that runs a huge share of discrete manufacturing got skipped, largely because it is specialized, unglamorous, and hard to learn from public data.
That hardness is exactly why purpose-built agents matter here more than anywhere. Infor platform knowledge lives in practitioners' heads and in twenty-year-old customization layers, not in web-scale training corpora. The gap between what a generic model knows about Mongoose, BODs, or M3 API programs and what a working consultant knows is enormous, and it is a gap you close with grounding, retrieval over real metadata, and encoded domain workflows, which is precisely what our 105 agents are.
Why Free Was the Right Call
The honest answer is that free is both a values decision and a strategy. The values part: the people who most need these tools are lean IT teams at mid-market manufacturers, three-person departments supporting 500 users, who will never get a six-figure AI platform budget approved. Charging per seat would have kept the tools away from exactly the users who prove them out. The strategic part: in a trust-starved market, working software is the only marketing that counts.
Free also compounds quality. Every team that runs the agents against a real environment finds edge cases, a Baan IVc quirk, a site-specific SyteLine event pattern, an M3 configuration we had not seen, and those findings flow back into the agent library. A paid, gated product would have a fraction of the exposure and therefore a fraction of the hardening. Eighteen months in, the agents are dramatically better because thousands of practitioner hours have hammered on them, none of which we had to pay for or could have simulated.
- The practitioners who most need AI tooling, lean mid-market IT teams, are the least able to buy platform subscriptions.
- Working software is the only credible marketing in a market burned by AI hype.
- Free distribution generates real-environment feedback that hardens the agents faster than any internal QA could.
- A large user base creates the standard: the agents become how work gets done in the ecosystem.
How Our Business Model Actually Works
Netray makes money the same way open-source infrastructure companies do: the software is free, and the outcomes are the product. Enterprises pay us to deploy the agents on their infrastructure, integrate them with their specific SyteLine, LN, M3, or ServiceMax environments, tune models on their data, and stand behind the result with SLAs. They also engage us for the AI-accelerated services the agents enable, migrations, upgrades, testing programs, integration rescue, delivered in half the traditional time because the agents do the archaeology.
This alignment is deliberate. A per-seat license model rewards us for lock-in; a services-on-free-software model rewards us only when deployments actually work. Our incentive is to make the free agents as capable as possible, because every capability increase makes the deployment engagement more valuable, not less. Roughly speaking, a typical enterprise engagement runs $50,000 to $250,000 depending on scope, against agent-driven savings that our clients measure in multiples of that in the first year.
- The agents are free forever; revenue comes from deployment, integration, tuning, and SLA-backed support.
- AI-accelerated services, migrations, upgrades, testing, are priced on outcomes the agents make possible.
- No per-seat licensing means no incentive for lock-in and no penalty for spreading the tools internally.
- Typical engagements run $50K-$250K against first-year measured savings that exceed the fee severalfold.
The On-Premise-First Bet
Every mainstream AI product in 2025 assumed your data would travel to someone else's cloud. For our market, that assumption is disqualifying. Aerospace and defense manufacturers under ITAR and CMMC, and any shop with customer flow-down confidentiality clauses, legally cannot send drawings, BOMs, and technical data to public LLM endpoints. We watched dozens of their AI pilots die in security review for exactly this reason, and concluded the bet was obvious: build for open-weight models running inside the customer's firewall.
The bet has aged well. Open-weight models crossed the capability threshold for ERP-grounded work, query generation, code analysis, document extraction, test authoring, while their hardware requirements fell to a single $30,000-to-$80,000 server. Meanwhile every headline about cloud AI data retention made on-prem an easier conversation. We think the end state for regulated-industry enterprise AI is inference at the data, not data at the inference, and the 105 agents are engineered for that world from the first line.
- ITAR, CMMC, and customer confidentiality clauses make cloud-only AI legally unusable for much of our market.
- Open-weight models now handle ERP-grounded tasks on a single affordable GPU server.
- On-prem inference removes per-token economics, enabling the heavy verification loops ERP work demands.
- Every agent is built cloud-optional: the same stack runs air-gapped, in a DMZ, or in a private cloud.
What's Next for the Agent Library
The library keeps growing where practitioners pull it: deeper CloudSuite Industrial coverage as more clients move off on-prem SyteLine versions, expanded M3 compliance agents coming out of our food and beverage work, and richer ION monitoring automation. We are also investing in agent-to-agent workflows, a migration agent that hands findings directly to the test case agent, for example, because the compound workflows are where the 50%-faster project numbers come from.
If you take one thing from this post, make it this: download the agents, run them against a sandbox, and judge them on your own environment. They are free precisely so that the evaluation costs you an afternoon instead of a procurement cycle. And if they prove out and you want them in production with someone accountable on the other end of the phone, that is the part Netray sells.
Key Takeaways
- 1The Infor ecosystem was skipped by the AI tooling wave because its knowledge is specialized and absent from public training data.
- 2Giving the 105 agents away for free maximizes adoption, feedback, and hardening while keeping incentives aligned with customer outcomes.
- 3Netray's revenue comes from deployment, tuning, SLAs, and AI-accelerated services, not per-seat licenses.
- 4On-premise-first is a bet that regulated-industry AI ends up as inference at the data, and ITAR/CMMC realities are proving it right.
Download the free agent library, run it against your sandbox this week, and see what purpose-built Infor AI actually feels like.
More Insights
Why 98% of Manufacturers Are Exploring AI But Only 20% Are Ready
The manufacturing sector is drowning in AI hype but starving for practical implementation. We analyzed 200+ manufacturers to understand the gap between AI exploration and AI readiness - and what separates the 20% who are actually deploying from the 80% still stuck in pilot purgatory.
AI AgentsThe Complete Guide to Infor SyteLine AI Agents
A deep dive into our 40+ purpose-built AI agents for CloudSuite Industrial (SyteLine). From IDO query optimization to BOD mapping automation, learn how each agent works and how to deploy them on your own infrastructure.