AI Adoption vs Trust Gap in SaaS Enterprise

Cloud-based AI agents are running into a wall nobody talks about: context starvation. Your files, your code, your workflows — they live on your machine. Every cloud-based agent has to reconstruct that context from scratch, every single time. It's slow, lossy, and fundamentally limited. This is exactly why we built Proficina as a desktop-first AI workspace. What we've learned building local-first agent infrastructure: → Desktop agents access your full file system, running processes, and local tools natively → Latency drops dramatically when inference happens on-device → Privacy concerns disappear when data never leaves your machine → Context persistence across sessions is trivial when state lives locally The chip competition heating up — with new players challenging the compute monopoly — is accelerating this trend. Cheaper local inference means desktop AI agents become viable for more workloads every quarter. Cloud AI had its moment for experimentation. But for serious productivity — the kind where an agent needs to understand your entire working context — desktop-first wins. We're shipping agent capabilities this quarter that only work because they run locally. No cloud round-trip. No context reconstruction. Just your full workspace, always available. Where do you think AI agents should live — cloud or desktop?

AI is exploding across sectors — but who's managing the software chaos it creates? In this fast-growth AI era, SAM isn't optional—it's your strategic guardrail. Generative AI tools, ML platforms, and agentic workflows multiply licenses overnight, often 3x faster than traditional apps. Without SAM, you're blind to sprawl, shadow IT, and compliance gaps that audits love to exploit. Smart IT leaders are leveraging SAM for: Real-time discovery of AI/ML software across cloud, SaaS, and on-prem. Usage analytics to reclaim unused GPU credits and idle API keys. Automated governance tying licenses to business value, not headcount. The result? Cost control amid 40%+ annual AI spend growth, plus audit readiness when vendors come knocking.

Repetitive tasks consume valuable time. AI agents change that. 🤖 Check out this infographic for an at-a-glance of variations in agent capabilities based on type and the different ways agents enhance productivity across an organization. @Microsoft 365

Software agents need more than a better harness. Once agents move from a laptop into cloud runtimes, the missing layer is a control plane: a central place to route models, govern MCP and secrets access, set budgets, and observe what agents are doing across an organization. In our latest post, we break the stack into three pillars: • Harness • Orchestrator • Control Plane We also explain why this matters now. Local agents are fine for experimentation, but as soon as agents run in containers, on schedules, or off third-party events, platform, security, and finance teams need visibility, policies, and audit trails. Some teams are already building this layer internally. We think it is becoming a real category for software agents. Blog: https://lnkd.in/g4_TuZCB Drafted with OpenHands AI for the OpenHands team.

For Brian Kinsella: Agents only work where people work. Unless they’re in the systems people are already using, agents can’t earn trust and scale real outcomes. Embedded agents are the only ones that actually get adopted. This is why architecture is everything.  #AI #IA #cloud #agents #pusspossible #manhattanassociates #future #technology

Every technology cycle has a lock-in trap. With AI, we're building one right now — and most organizations don't see it coming. Here's how it works: you buy an AI license from your favorite cloud provider. It connects to your data. Your employees start relying on it. Your workflows get built around its interface, its capabilities, its pricing. Then the price goes up. Or the model changes. Or the vendor decides to discontinue the tier you're on. And now you're stuck. Because your data, your workflows, and your team are all adapted to that one vendor's system. This happened with enterprise software. It happened with cloud storage. It's happening now with AI tools across every industry. The antidote is straightforward but it requires intentionality: own your AI infrastructure. Choose platforms where you hold the code, not just a license. Insist on LLM-agnostic architecture so you can switch models as better or cheaper ones emerge. Treat AI as institutional infrastructure — not a subscription service. The decisions you make now about AI vendor relationships will define your cost structure and flexibility for the next decade. Lock-in doesn't happen overnight. It happens one workflow at a time.

AI is no longer reserved for large corporations or massive budgets. Today, it is practical, accessible, and capable of solving real business problems, often at a cost lower than a single hire. In this blog, we break down why small and medium businesses actually have the advantage in the AI era, where the fastest wins are, and why most companies get implementation wrong. The focus is not on tools, but on processes that drive real results. If you want to understand how to start without a big budget and where to find the highest ROI, read the full blog here: https://lnkd.in/d5wee7Uj

If governance is automated, it’s easy. If it’s easy, it gets done. If it gets done, it gets implemented. Most AI initiatives don’t fail on models. They fail on governance. Manual governance doesn’t scale. Workshops don’t scale. Partner‑heavy setups don’t scale. Production needs defaults. That’s why Governance as Code matters. When governance is: 🟢 Automated 🟢 Enforced by design 🟢 Deployed like infrastructure Then AI can move from pilot to production. Not someday. Not after another framework. Now. We have it. #GovernanceAsCode #AIGovernance #AIinProduction #EnterpriseAI #PublicSectorIT #Azure #PlatformThinking #FromPilotToProduction

96% of enterprises are deploying AI agents. 94% of those same enterprises are worried about losing control of them. That's not a coincidence. That's what happens when you build AI the same way you bought SaaS — from a vendor, on their infrastructure, with no code access and no audit trail. Every agent running on someone else's stack is a dependency you don't fully control. When the vendor changes pricing, you pay. When they change capabilities, you adapt. When you want to know what the agent actually did, you file a ticket. The organizations building durable AI capability aren't buying more licenses. They're owning the infrastructure.

“While a collapse in the SaaS business model seems improbable, recent advances in AI are a paradigm shift in how organizations develop and procure software capabilities.” https://lnkd.in/ej4GbDJq