Sheri Hinish, Founder of Supply Chain Queen® and CEO of Supply Chain Revolution Global, hosts John Lane, Vice President of IT Application Solutions and Implementations at Flex, on the Supply Chain Revolution® podcast. Enterprise AI has hit an inflection point, and it all starts with your processes, not your models. John explains how Flex, an Agentforce Operations customer, is using Salesforce’s Agentforce Operations, powered by Regrello, to run processes with AI from the ground up across its core back-office operations. The result is automated processing for 95% of work order approvals and roughly 70% of third-party risk assessment activities. This episode is a paid partnership with Salesforce (#ad).
Why Enterprise AI Value Starts With Your Processes, Not Your Models
The market has moved past the hype. The question every executive now asks is how quickly AI can deliver measurable value, and most organizations are discovering that the answer has little to do with model horsepower. My perspective is that the answer starts with your processes. Most enterprises still run their back offices on decades of manual workarounds: approvals chased by email, data rekeyed between systems, and handoffs nobody owns. When AI is bolted onto those workarounds, it does not fix them. It simply moves the same friction faster.
Flex offers a practitioner’s view of the alternative. In this conversation, John Lane walks through how a global manufacturer decided where to deploy agentic AI, what happens inside a workflow when work stalls, how legacy systems connect without a costly migration, and where human judgment remains essential.
Why Did Flex Bring Agentic AI Into Back-Office Operations?
Flex is a contract manufacturer. It builds products for other companies, along with some of its own, across a significant global manufacturing footprint. Behind those factories sits a very large back-office organization that supports them centrally. John described that environment as full of repetitive, high-volume work, from financial transactions to engineering change orders, which is exactly where an agentic approach adds significant value.
The goal was not to experiment with AI for its own sake. The goal was to bring measurable efficiency to the work that keeps a global manufacturing network running, and to do it in partnership with Salesforce through Agentforce Operations.
Why Should Process Come Before AI?
When I asked John how process mapping shaped where Flex chose to deploy AI, his answer reframed the question. Flex already understood that repeating a manufacturing process or a supporting back-office process at global scale depends on strong standard processes. What the agentic era reinforced is that process engineering becomes even more critical once agents enter the picture.
“Adding AI into a poor process just gets us someplace, the wrong place faster.”
John Lane, Vice President, IT Application Solutions and Implementations, FlexJohn described a clear sequence. Flex first defines the end-to-end process, then digitizes it, and only then layers agents on top. The temptation, he noted, is to jump straight to an agent, and teams that give in to it end up with pocket solutions rather than results at scale. The real work is understanding how to orchestrate the process and deciding where an agent, or several agents working together, actually makes sense.
John described a pragmatic approach that I admire: prove it before you build it. Leaders who map how work actually runs, measure it against a real baseline, and then decide where agents act are the ones who turn AI ambition into operational value.
How Does Agentforce Operations Build Back-Office Workflows in Minutes, Not Months?
John pointed to the simplicity of the framework as the differentiator. The individual agentic capabilities, such as creating, analyzing, and researching, matter, but the real value comes from sequencing them inside a clear process wrapper. Some steps are straightforward approvals. Between two steps, however, the workflow may need to analyze a document, extract information, and present it to the next person who must act.
According to John, assembling a workflow like that now takes minutes to hours, compared with the days, weeks, or months it used to require. He also highlighted the ability to call out to other systems, whether to extract data or push it back in, as a major accelerator for his team.
Related reading: How AI Agents Are Different from Traditional Supply Chain Software
How Can You Connect Legacy Systems to AI Without a Platform Migration?
For decades, connecting older systems meant slow, expensive, and brittle integration work. Many leaders know that cost as the integration tax, and nearly every large enterprise pays it. Flex is no exception. As a large manufacturer that has acquired companies over the years, it still runs some applications that are far from ideal and will not be replaced in the near term.
Flex took part in beta testing for the Systems Integration Agent within Agentforce Operations. John explained that the capability simulates a user’s interaction with an application: it reads the interface, picks up the data, and understands it. Flex proved the approach on a true legacy application, and the agent built an MCP layer for that application at the front end.
“It’s got a double benefit.”
John Lane, on the Systems Integration AgentThe first benefit is speed, because the connection happens far faster than a traditional integration project. The second is reuse, because the MCP layer the agent produces becomes an integration asset that Flex can use in more traditional integration methods as well. John admitted that when the capability was first described to him months earlier, he laughed and asked to see it once it worked. It does, and he called it a big step forward for legacy connectivity.
Where Does the Human Stay in the Loop With Agentic AI?
Work order approvals show how the balance between agents and people plays out. Once Flex finishes producing a run of units and consumes the material, every work order also requires a financial close. The core question is whether the work order cost more than it should have. Agentforce Operations draws on data from Flex’s ERP and manufacturing execution systems, applies criteria that can differ by product, checks the variance against the allowed tolerance, and automatically approves the work orders that fall within it.
People step in when a business decision is required. If a variance is tied to a particular customer or event, part of that cost may be claimable, which makes it a revenue question rather than a routine approval. In those cases, an account program manager works directly with the customer to agree on the outcome. John noted that Flex hopes to bring customers into the loop over time as well.
John made a point that every operations leader should hear. Before automation, Flex had a human in the loop for everything, and because no company can staff enough people to review every work order, some variances were simply accepted. Automation now applies consistent logic across the work and learns from it, which stops what John called a silent leaky bucket of loss.
Third-party risk assessment follows the same pattern with a different decision owner. When Flex onboards a software vendor, its CISO applies a strict security standard. The process once required extensive back-and-forth with each vendor about documentation such as SOC 2 reports. Today, vendors submit their documents by email, the agent extracts and assesses them, and the CISO receives a prepared assessment report with a recommendation. Given how critical security is, John was clear that a human will remain in that loop at all times. What changed is that his security experts no longer spend hours on purely administrative work.
How Is Flex Measuring the Impact of Agentic AI?
On the business side, John described the value in terms of throughput and upside that ultimately reaches operating profit. On the IT side, the bigger shift is speed. Because his team can now design and run agentic workflows in a much shorter time, it can pursue use cases that were never worth the effort before.
The obvious, high-volume opportunities come first. Once those are in place, smaller use cases that might each save the effort of a handful of people become viable, and they add up quickly. John was candid that the industry is still at a relatively early stage. He sees the internal skills of his own team, combined with tools like Agentforce Operations, as the game changer for how much value Flex can capture.
Agentforce Operations “orchestrates end-to-end workflows across our core systems, enabling teams to focus their expertise where it adds the most value.”
John Lane, Vice President, IT Application Solutions and Implementations, FlexHow Do Executives Turn AI Hype Into Measurable Business Value?
John’s advice was refreshingly grounded. There is an element of truth to the hype, he said, because the tools really do these things. The leaders who capture value, however, always come back to two fundamentals: understanding the process and understanding the data. A strong idea for improving a process fails without a clean, available data set that matches it.
“One doesn’t come without the other.”
John Lane, on process and dataFor IT leaders, many familiar challenges remain, including getting data out of the ERP correctly and making it accessible. John expects a more fully orchestrated, agentic environment to change that over time, allowing agents to tap into data sets without moving the data as much as enterprises have traditionally had to.
Key Takeaways for Supply Chain and Operations Leaders
AI value depends on process discipline. Model horsepower matters less than a clear, standard, end-to-end process. Adding AI to a poor process only gets you to the wrong place faster.
Sequence the work. Flex engineers the process, digitizes it, and then adds agents, which allows solutions to scale beyond isolated pockets.
The integration tax is optional. The Systems Integration Agent connects legacy applications without a platform migration and produces a reusable MCP layer for future integrations.
Design the human role deliberately. Agents handle the routine approvals, while people make the decisions that involve customers, revenue, and security.
Data is the other half of the equation. Clean, accessible data must match the process you are trying to improve.
Why the Back Office Is the Next Frontier for Enterprise AI
The front office got its data and AI discipline first. The back office is clearly next. Flex shows what that shift looks like when it is built on the process rather than layered on top of it: workflows assembled in minutes instead of months, legacy systems connected without a migration, and people freed to focus their expertise where it adds the most value.
Sheri Hinish is the Founder of Supply Chain Queen® and CEO of Supply Chain Revolution Global. She hosts the Supply Chain Revolution® podcast. Previously she served as EY’s Global Consulting Leader for Sustainability, Technology, and Ecosystems and as a Senior Executive at IBM, advising Fortune 500 companies and governments on supply chain transformation, AI-enabled operations, sustainability, and innovation. She holds advanced degrees from Harvard University and Rutgers University and is recognized as the #1 Global Supply Chain Influencer. Agentic AI, AI-enabled operations, and the redesign of supply chain and back-office processes are central themes of her advisory and media work.
Learn More
Agentforce Operations: Explore Agentforce Operations, powered by Regrello | Flex: flex.com | Supply Chain Queen: supplychainqueen.com
If this conversation was valuable, listen to the full episode, share it with an operations leader who is ready to stop paying the integration tax, and follow Supply Chain Revolution on YouTube, Apple Podcasts, and Spotify. Until next time, keep leading the revolution.
Sheri Hinish is the Founder of Supply Chain Queen® and CEO of Supply Chain Revolution Global LLC, a supply chain thought leadership, advisory, media, and community platform. Previously she served as EY’s Global Consulting Leader for Sustainability, Technology, and Ecosystems and as a Senior Executive at IBM, advising Fortune 500 companies and governments on supply chain transformation, AI-enabled operations, sustainability, circular economy, and climate strategy. She holds advanced degrees from Harvard University and Rutgers University and is recognized as the #1 Global Supply Chain Influencer, a Top 250 Global Leader in Sustainability, Top 100 Women in Supply Chain and Technology, and Top 100 B2B Influencer in North America. She hosts the Supply Chain Revolution® podcast and speaks globally at COP, NY Climate Week, CES, and major industry events.