AI Adoption Is Accelerating.Most of the Hard Parts Are Still Hard.
Enterprise AI budgets are up. Timelines are compressed. And yet inside most organisations, the honest conversation sounds something like: “We’ve got pilots. We’re just not sure they’re moving the business.” We see this across the Predixion ecosystem every day. Here’s what the full picture actually looks like.
The blockers nobody budgets for.
A company announces an AI initiative. A vendor gets selected. A pilot runs and looks promising. Then someone asks how to scale it — and the room goes quiet. The problems that stop enterprises cold are rarely about the AI itself. They’re about data locked in silos, legacy systems that weren’t built to connect to anything new, and a workforce that’s heard enough transformation announcements to be quietly sceptical about this one too. These don’t get solved by buying a better model.
ERP is where it gets real — and where most organisations get stuck.
One of the largest clusters in the Predixion ecosystem is business software: ERPs, CRMs, and the platforms that run day-to-day operations - there are over a dozen vendors in the ecosystem working across these platforms, which tells you something about where enterprise demand is actually sitting.
The most significant shift we’re seeing is AI moving from a reporting layer to a decision layer inside these systems. One vendor has rebuilt their ERP around this idea entirely — instead of dashboards, an agentic system that monitors financial flows, flags anomalies, and proposes actions before anyone has opened the application. Others are deploying AI agents inside existing SAP and Salesforce environments to automate accounts payable, vendor onboarding, and approval workflows that previously ate hours of manual effort every week. The measurable results show up in processing times and error rates, not in slide decks.
Supply chain AI is further along than the conversation suggests.
One area the broader AI narrative consistently underestimates is supply chain. In the ecosystem we work with vendors who’ve built AI platforms specifically for integrated business planning — combining demand forecasting, risk modelling, and logistics optimisation in a single operational view. One applies digital twin technology to simulate disruptions before they happen, giving procurement teams lead time to respond rather than react. Another optimises freight networks in real time, cutting empty miles on every load. These aren’t AI features bolted onto existing tools. They’re purpose-built, and they’re in production.
Strategy before tools. The vendors helping enterprises get the sequence right.
Something we see repeatedly: organisations that invest in AI capability before they’ve figured out where AI actually belongs in their business. The ecosystem includes a meaningful cluster of data and AI strategy consultancies doing the harder upstream work — use case prioritisation, GenAI roadmaps, data maturity assessments, and AI value realisation frameworks. This work isn’t glamorous, but it’s often what separates organisations that get ROI from those that accumulate a graveyard of promising pilots.
Custom agents, GenAI, and the vertical-specific capability that matters most.
Beyond the platform ecosystem, the ecosystem covers vendors building custom AI from the ground up: LLM fine-tuning, MCP server development, agentic workflows for complex multi-step processes, computer vision for manufacturing inspection, and conversational AI for banking and collections. What’s striking is how vertical-specific the best work has become. A talent intelligence platform modelling behavioural fit and predicting retention. An AR/VR training platform replacing paper procedures with immersive simulations in pharma and industrial settings. An AI-powered OTT workflow engine automating subtitle generation, live-to-VOD clipping, and content scheduling. Generic AI is everywhere. Specialist capability built for a specific workflow is considerably rarer and considerably more valuable.
Security is the quiet blocker most AI programmes hit too late.
AI systems introduce new attack surfaces, pull data across systems in ways that create unexpected exposure, and generate outputs that can be manipulated if the environment isn’t hardened properly. The ecosystem has real depth here — AI-powered continuous exposure management, VAPT across web, mobile, cloud, and IoT environments, managed SOC services, and vCISO-level governance. One vendor runs AI agents that continuously discover and auto-remediate vulnerabilities across production environments. For organisations moving AI into core business systems, this kind of always-on posture isn’t optional.
People are still the variable that determines whether any of this works.
An enterprise can deploy a well-built AI system and watch it sit unused for months because the workforce wasn’t brought along. The ecosystem covers this too: AI-native learning platforms using role-play simulation rather than slides, talent intelligence tools that map organisational capability against what AI adoption actually demands, and a human-AI hybrid mental health platform that addresses the anxiety large-scale change quietly generates. These aren’t soft additions. They’re often what makes the difference between a deployment that sticks and one that quietly gets abandoned.
The enterprises closing the gap between AI investment and AI impact share one thing: they stopped treating AI as a technology project and started treating it as a business redesign. The sequence matters. Strategy before tools, infrastructure before agents, people before go-live.