Insights
Tracking What Matters
Key announcements, milestones, and moments that matter—from Innovative Solutions and our partners.
This includes company news, AWS recognitions, product launches, and updates that reflect how we’re growing and where we’re headed.
FILTER INSIGHTS
AI agents can now log into systems and act for a person: updating records, running commands, deleting files. That raises the stakes on their mistakes, because an agent’s error happens inside the systems it can reach. This post covers three things to settle before giving an agent access to real accounts: what it should be able to reach, what actually stops it, and who answers for what it does.
Agentic coding tools have made it realistic for companies to build software they once had to buy, which puts the build-vs-buy question back on the table. But a fast build is only one part of the job. This post looks at why the decision is worth revisiting, what happens when a quick internal tool becomes something the business depends on, and where the real effort in software now sits.
No single choice carries your whole system.Businesses want one model, one decision, one person to hold the whole answer. AI rarely lets that hold for long. No single model fits every task, which is why the best systems route across several instead of betting on one. A model that performs well today can quietly fall behind within months, whether anyone is watching or not. And the reasoning behind a decision doesn't travel on its own; it leaves the building the day the person who understood it does.
In Episode 17 of AI Unplugged: AI Gone Rogue: Security, Slop & the Cost of Intelligence, our team unpacks what happens once AI capability stops being the open question. An agent reroutes around a system's security to finish its task. A frontier model's price drops overnight. In both cases, the problem isn't whether AI can do something, it's who's responsible, and who actually benefits, once it does.
If you passed on an AI project because the cost didn't add up, it may be time to revisit that decision. Routine AI work has gotten meaningfully cheaper, but the most advanced models haven't. Here's what's driving the shift and how to figure out which parts of your project belong on which pricing tier.
Professional services have priced work by the hour for decades because the hour was the one thing both sides could measure and agree on. AI is breaking the assumption that pricing depends on: that hours worked and value delivered move together. What replaces it isn't settled yet.
No single choice carries your whole system.Businesses want one model, one decision, one person to hold the whole answer. AI rarely lets that hold for long. No single model fits every task, which is why the best systems route across several instead of betting on one. A model that performs well today can quietly fall behind within months, whether anyone is watching or not. And the reasoning behind a decision doesn't travel on its own; it leaves the building the day the person who understood it does.
AI systems can lose effectiveness as models are deprecated, standards evolve, and data shifts, even when the original build was excellent. This article makes the case that AI should be treated as a maintained capability rather than a finished project, and shows what that discipline looks like in practice.
Every time an AI project changes hands, valuable business and technical understanding gets lost. This article explores how Forward Deployed Engineering preserves that context so knowledge and value can build over time.