The 3 Pillars That Make AI Actually Work for Manufacturers
By Christian Taylor (Solution Architect) & Shannan Robillard (Salesforce Practice Director)
Research from MIT found that 95% of all AI projects result in zero return. In most cases, the issue isn’t the technology, but instead everything that needs to happen around it. Oftentimes, we see data that isn’t aligned, teams that don’t change how they work, and systems aren’t set up to support continuous improvement.
That’s why we say the go live on any new AI-related implementation is only the starting line. The long-term value comes from the foundation you build next.
Why do so many manufacturers struggle to turn AI pilots into real business outcomes, while companies like Kawasaki Engines and Luminator accelerate value instead? The difference almost always comes down to the foundation you put in place.
To avoid becoming another stalled AI pilot, manufacturers need three things working together:
- A strong Data Cloud foundation
- Change management built specifically for AI adoption
- A continuous improvement model that keeps agents accurate and relevant as the business evolves
When these pillars are there, AI has the conditions it needs to succeed. Here’s what the companies in the 5% are doing differently.
Pillar 1: A Clear Data Strategy Anchored in Data Cloud
Most manufacturers operate across a mix of systems, including ERPs, Salesforce, service platforms, warranty tools, and spreadsheets, not to mention data that lives outside of IT’s view altogether. That fragmentation makes it difficult for AI to reason across the business, and it makes it hard for teams to trust outputs.
Data Cloud changes this by creating a single, unified data layer that brings siloed information together. It does this by bringing structure and consistency to the data that powers sales, service, operations, and (eventually) AI.
When that foundation is in place, reporting becomes consistent, teams can make faster decisions more efficiently, leaders get one version of the truth, and AI agents finally have reliable inputs instead of stitched-together fields that produce inconsistent answers.
A recent ForeFront engagement shows what this looks like in practice. One building materials manufacturer ran four ERPs and four separate business units. There was no easy way to understand customers, inventory movement, or product performance across the organization. Once Data Cloud unified those systems into a single customer and product view, every downstream workflow improved: Forecasting stabilized, reporting is accurate, and the business is now equipped to explore AI and Agentforce use cases without worrying that “bad data” will undermine adoption.
Think of Data Cloud as the engine under the hood that transforms scattered data into the fuel that powers a smarter more efficient business. It gives AI agents the intellectual power to streamline workflows and elevate user experiences. Once that is in place, the next goal is to inspire teams to embrace the technology in a new way.
Pillar 2: Organizational Change Management (OCM) Designed Specifically for AI Adoption
As Shannan often says, “Technology adoption is a change initiative, not a hand-off.” This is especially true with AI agents.
AI usage isn’t mandatory for teams to do their jobs. A sales rep can still type directly into an opportunity, a customer service rep can still search manually, and people can still export spreadsheets. This means adoption doesn’t happen automatically — it has to be nurtured, explained, and reinforced.
When AI pilots fail, it usually has little to do with the data model behind the scenes. It’s almost always the surrounding environment: data that isn’t unified, processes that aren’t aligned, and workflows that don’t match how people do their work every day. Users don’t know what the agent can do, or why they should trust it, so they revert to familiar habits.
Effective OCM meets teams where they are. It shows people exactly how AI supports the tasks they do today and makes those tasks easier. It builds confidence through repetition and reinforcement. Ultimately, it turns a new tool into something teams actually want to use because it helps them do their work more effectively.
At ForeFront, this work starts with understanding how the business really operates, including how sales reps enter orders, how service techs capture failures, how product data flows across regions, and so on. From there, we align the data foundation to support those processes, not compete with them. We build enablement around real workflows, not abstract demos, and we develop internal champions who can model new behaviors and give teams a trusted point of reference.
The goal isn’t to force new behavior, it’s to show how AI removes friction from work people already need to do. When people experience the difference for themselves, adoption tends to follow.
We saw this play out with Luminator. Once their customer service team understood how the agent supported their work, they were eager to use it to automate responses to routine inquiries and free up time to focus on higher-value tasks. The efficiency boost happened because adoption was anchored in real workflows and pain points, not abstract training.
How did ForeFront help Luminator make AI adoption stick? By anchoring every training and workflow change to real service scenarios, not theoretical use cases. That’s the difference between a pilot that fades and a program that scales.
Pillar 3: Continuous Improvement (Not One-and-Done)
One of the most common reasons AI projects stall and fail to deliver value is that teams try to deliver everything all at once. They try to build a fully mature agent on day one, with the aim of covering every workflow and every edge case. When the agent inevitably gets something wrong, users lose trust and the program loses momentum.
A better approach is to think big, start small, deliver quickly, and build over time. An AI agent’s accuracy improves as data matures, people adjust workflows, and teams provide feedback.
Kawasaki Engines saw the value of this approach firsthand. They needed a partner who could provide the winning combination required for adoption to grow over time: both technical integration capabilities and alignment with the business.
ForeFront integrated Salesforce with their legacy LANSA ERP, which has given them a modern B2B platform that supports dealer operations across the business. We’ve also stayed engaged as a strategic partner to help them align the latest technology with day-to-day priorities. That level of guidance is what enables companies to expand Salesforce adoption over time and continue improving as their needs evolve.
This kind of disciplined, phased approach works because it mirrors how manufacturers drive operational improvement: steady progress, consistent tuning, and incremental gains that build on each other. When teams have a foundation they trust and a roadmap they can follow, AI stops feeling experimental and instead becomes part of how the business operates.
Where Agentforce Fits In
Data Cloud serves as the context engine, providing Agents with the unified, relevant data needed to understand the world they operate in. To recap our three pillars of AI excellence:
- Data Cloud gives manufacturers a unified view of their business.
- OCM helps people use the tools built on top of that view.
- Continuous improvement keeps those tools aligned with the reality of the operation.
Agentforce sits at the intersection of all three.
With a reliable data foundation, Agentforce agents can answer questions accurately, retrieve information from multiple systems without manual digging, and support decisions in real time. For sales, that might mean instant access to open quotes, inventory availability, order history, or pricing. For service, it might mean faster troubleshooting and fewer repeat visits. For leaders, it means they can spot patterns across customers, product lines, and regions without reconciling competing reports.
When data is consistent and workflows are clear, AI becomes a trustworthy and dependable sidekick.
ForeFront’s Role as a Continuous Improvement Partner
Even the best-designed systems become less effective over time if no one maintains them. Business processes shift, product lines change, new data sources come online, and the people using the system often aren’t the same people who were trained at go live. Without ongoing alignment, Data Cloud loses accuracy and AI agents lose context, which slowly erodes trust and adoption.
This is where the change-management work becomes most critical, and where ForeFront’s AI-first delivery approach helps teams adopt AI in a way that feels natural instead of disruptive. It’s the same “meet you where you are” model that Tom Dennis (Technical Delivery Director at ForeFront) outlined in
Because of this, ForeFront doesn’t view go live as the conclusion of the project. Go live is the point where ongoing partnership matters most. We stay close to our customers’ teams to tune models, refine identity rules, adjust prompts, bring new sources into Data Cloud, and guide the natural evolution of AI use cases. Our measure of success isn’t the launch date; it’s how the organization operates months later and whether the system is still supporting the way the business runs.
We look for signs that the program is maturing, like cleaner data, more consistent workflows, higher adoption, and insights that shape decisions instead of sitting in a dashboard. When those things are happening, it’s signals that the foundation is healthy and value will compound over time. That’s the mark of a sustainable AI program, and it’s the outcome we help manufacturers reach.
About the Authors
Shannan Robillard is a Practice Director at ForeFront, bringing more than 10 years of consulting and program leadership to enterprise Salesforce initiatives. She specializes in organizational change, user adoption, and helping teams turn complex technology into everyday tools that work.
Christian Taylor is a Solution Architect at ForeFront with deep experience in Data Cloud, MuleSoft, and multi-system integrations. He builds unified data architectures across ERP, CRM, and service systems, giving organizations the foundation they need for accurate reporting, automation, and AI.
