From my seat – inside deals, renewals, and delivery conversations every week – the story from early 2025 to early 2026 is simple: last year’s “innovation pilots” just turned into this year’s invoices, and a lot of leaders are realizing the math doesn’t work under old ways of buying and running enterprise software.
This isn’t really a “new tech” story; it’s a story about operating models colliding with AI and SaaS economics in life sciences, CPG, and manufacturing.
1. Last year’s pilots became this year’s bills
In 2025, many enterprises treated AI and new SaaS tools as low-risk experiments – run a few pilots, spin up some licenses, bolt on another “copilot,” worry about the integration later.
Now it’s 2026, renewal season is here, usage is all over the place, and CFOs are staring at stacked line items where every vendor quietly added an “AI uplift” to their pricing model.
A few patterns I’m seeing everywhere:
- SaaS and AI are being used to justify price hikes, not just value. Many vendors added AI features and used that as cover to raise subscription costs, even when the AI capabilities are basic and adoption is low.
- Tool sprawl got worse, not better. The average enterprise is now managing hundreds of SaaS apps, with overlapping AI features and no coherent view of who’s using what or why.
- Cloud and data costs are catching up with AI enthusiasm. Training, inference, and data movement for AI workloads are driving unexpected infrastructure bills on top of application subscriptions.
If 2025 was the year of “let’s try it,” 2026 is the year of “what are we actually keeping – and what are we willing to cut.”
2. AI went from toy to infrastructure
On the upside, AI is no longer just a demo of a chatbot talking to your ERP.
The shift I see between last year and now is that AI has become infrastructure: embedded in forecasting, quality, planning, and development workflows rather than sitting off to the side.
Across supply chain, BTP-style platform work, and process tools, a few things stand out:
- Domain-specific AI is replacing generic “ask me anything” bots. Models tuned for planning, quality inspection, or clinical data are beating general-purpose bots on accuracy and trust, especially in regulated industries like life sciences.
- Agentic AI is starting to do actual work, not just generate suggestions. In areas like observability and security, AI agents are autonomously triaging incidents and taking actions, which is a preview of what’s coming for more business processes.
- Development teams are quietly rebuilding their workflows around AI. Code generation, test creation, and documentation are increasingly AI-augmented by default, which changes the economics of test automation and integration work.
In other words, AI stopped being “that chatbot our CIO showed on a town hall” and started becoming part of the plumbing – especially where there’s a lot of structured process data to learn from.
3. Your operating model is now the bottleneck
Here’s the uncomfortable part: the main constraint in 2026 is no longer “do we have AI” – it’s “are we organized to use it at scale without chaos.”
Most of the organizations I work with still fund and govern technology like it’s 2012: large projects, annual budgeting, siloed IT vs. business, and change managed as an afterthought.
Meanwhile, the leading edge is moving in a different direction:
- “AI-native” product teams are becoming the baseline. Cross-functional teams that own a business outcome (like perfect order fulfillment or batch release cycle time) and have embedded AI/data capabilities are beating traditional project teams on speed and learning.
- Operating models are being rewritten around automation and agents. Some firms are explicitly designing processes assuming AI handles a first pass for routine decisions, and humans focus on exceptions and strategy.
- Governance is shifting from gatekeeping to guardrails. Instead of committees approving every AI experiment, leading organizations define data, security, and compliance guardrails – then let teams move faster within those boundaries.
If your IT and business teams are still organized around projects, cost centers, and tickets, you will squeeze almost no value out of AI, no matter what platforms or vendors you pick.
4. What’s happening to the legacy stack (and vendors)
Another underreported shift from 2025 to 2026 is what’s happening under the surface to legacy stacks and long-time vendors.
From what I’m seeing with SAP and non-SAP landscapes alike:
- Platforms are winning; disconnected point solutions are under pressure. Enterprises want fewer platforms that can host AI, workflow, and integration consistently, rather than a zoo of niche tools that each bolt on their own “AI.”
- Open and API-first tools are quietly gaining ground. Open-source and open-core tools for AI deployment, MLOps, and automation are attractive because they provide flexibility and help avoid vendor lock-in, especially as AI usage patterns are still evolving.
- Legacy vendors are using AI to defend, not just disrupt. Many large providers are shipping “AI inside” and analytics features mainly to keep customers in the ecosystem and justify higher prices, not necessarily to rethink how the work gets done.
SAP clearly fits into this story as one of the big platforms: there’s real innovation in areas like business AI, process intelligence, and integration, but the value depends almost entirely on how you integrate those capabilities into your operating model and your broader toolchain.
5. What leaders can actually do this quarter
If you’re a VP or director in IT, supply chain, manufacturing, or a business function, the question isn’t “how do I keep up with AI” – it’s “how do I stop paying for noise and start designing for outcomes.”
Here are concrete moves I’d be making in 2026:
1. Treat SaaS and AI as a portfolio, not a collection of renewals
- Build a simple health view: where are you paying for AI/SaaS, what’s adoption, what business outcome is it tied to.
- Cut or consolidate tools that duplicate capabilities or have low engagement, especially where a core platform can reasonably absorb that function.
2. Redesign a few critical workflows as “AI-first” experiments
- Pick one or two processes in supply chain, quality, or finance where there is enough data and clear rules (e.g., demand planning, exception handling, test execution).
- Explicitly define what the AI will do end-to-end (ingest, recommend, act, escalate) and what the human role is, then measure before/after on cycle time, errors, and cost.
3. Move one or two teams toward an AI-native product operating model
- Create small, cross-functional teams that own an outcome and have embedded data/AI capability instead of relying on a central “AI team” for everything.
- Fund them as products with roadmaps and success measures, not as one-off projects that die after go-live.
4. Put guardrails around AI – then get out of the way
- Define clear standards for data privacy, model usage, prompt safety, and regulatory compliance, especially in life sciences and regulated manufacturing.
- Once guardrails are in place, encourage teams to aggressively automate tasks and decisions within those boundaries instead of waiting for central approval on every use case.
5. Be ruthless about “AI theater”
- If a vendor can’t show you how their AI changes a specific KPI in a process you care about, assume it’s a feature for the demo, not for your P&L.
- Internally, stop celebrating pilots and start celebrating decommissioned legacy tasks and tools – that’s the real signal that AI is doing work, not just making slides more interesting.
From 2025 to 2026, the game changed less in the headlines and more in the plumbing: AI got real enough to be useful, SaaS and cloud costs got high enough to hurt, and the enterprises that win will be the ones that change how they buy, structure teams, and design work – not just the ones that buy the most “intelligent” software.
If you’re feeling both excited and slightly sick to your stomach about your AI and SaaS spend right now, you’re not behind – you’re just in the phase where the hype passes and the operating model work begins.
