As organizations look forward to optimizing workflows through AI, many are finding that legacy tech is holding them back.
Across every industry and segment, organizations know that they need to integrate AI and agents into their everyday workflows. In fact, research shows that 93% of CIOs plan to introduce autonomous agents by 2028. For most others, it’s on the roadmap. Unfortunately, legacy tech is holding many companies back.
Why AI initiatives fail to scale
Legacy modernization is easier said than done. These are the top six challenges our customers face when trying to migrate from legacy integrations and systems:
- Taking on too much, AKA “big bang” migration: Moving everything at once is chaotic and high risk. People are naturally fearful. If something breaks, everything stops.
- Jumping in without a strategic plan and partner: We like this to drive cross-country without a map. Funding requests demand clear ROI, so a clear plan for modernization and an experienced partner can provide the necessary insights and keep integrations from creating more problems than you started with.
- Thinking short term and choosing the wrong tools for the job: It’s tempting to seek a quick fix or save a few dollars up front. However, without tools that will scale, today’s fixes won’t meet tomorrow’s needs. You’ll just kick challenges and costs down the road.
- Overview security and emerging vulnerabilities: Any IT change creates new security risks. Neglecting security integration and governance opens the door to attacks.
- Moving forward without defined standards: This can lead to the accumulation of technical debt.
- Neglecting change management: Success will be hard to achieve if you don’t gain buy-in from and upskill those who will be responsible for making your new systems work.
Kurt Anderson, Managing Director and API Transformation Leader at Deloitte Consulting LLP, joined us in June 2025 for a webinar discussing how companies can mitigate these challenges while modernizing:
“Everyone wants to have an AI strategy and be a part of the solution, but moving past the pilot stage is often tricky. Getting new tech to scale runs into hurdles when you don’t have clear business objectives or the right data at the point of the LLM, or if you overestimate how quickly you can deliver.”
The good news is that your organization can avert these problems with well-defined goals, clear data, a thoughtful API strategy, and a systematic approach to setting stakeholder expectations.
A strong API-led foundation
Organizations that have already invested in an API-led architecture will find themselves especially well-prepared to adapt to the needs of AI.
For companies in which AI and integration are both emerging initiatives, the foundational API layer and core AI capabilities will need to be put in place at the same time and grow together. After all, AI agents are only as smart as the data they can consume. A composable foundation will allow you to deliver the business value stakeholders demand.
While the average large enterprise manages over 900 discrete applications, not all will require integration. At MuleSoft, we apply an 80/20 rule, asking:
- Which platforms drive the business?
- What data does the LLM actually need?
- What major transactions deliver value?
Knowing which apps are necessary but unavailable allows you to prioritize. Start with three or four core applications, select the next few to activate, and continue down the line. This technique also makes it easier to see when it’s time to stop.
A “yes to both” attitude towards transformation
CIOs face a dual mandate to support the legacy architecture that runs the business while embracing new technologies. A bimodal strategy allows you to excel at both.
Anderson elaborated: “You have to keep the business running, with the right teams and processes, while looking for efficiencies. Perhaps you can use AI tools that provide the same quality output at a lower cost. That’s going to free up funding that you can use to introduce new capabilities and more agility. You get a virtuous feedback loop – taking costs out of the legacy system and getting more efficient with modern options. By the second or third initiative, you can see how the impact is compounded.”
This can get the flywheel spinning faster, especially when paired with the reusable nature of modern API-led architectures. It also underpins a strategy for reducing technical debt.
As you prioritize workflows, you can consider moving them to a native cloud environment or best-of-breed platform to improve scalability, security, transparency, and performance. Just because your technology is old doesn’t mean that the process for modernizing it needs to be. AI-enabled migration can help you understand the legacy portfolio, optimize platforms, take out keystrokes, and advance the way that you test and deploy.
Balancing today’s processes and tomorrow’s promises
Business-led innovation that capitalizes on the potential of agentic AI while satisfying the interests of customers and stakeholders is possible, with measured, purposeful processes in place. A composable, API-led architecture and a bimodal approach to supporting and upgrading legacy tech can move your organization forward without disruption.
Enterprises that have adopted these methods have accelerated migration times by up to 35% while reducing associated costs by as much as 50% – all while de-risking the complete effort.
The right partner and tools are key – starting with a well-informed strategy and a reliable migration platform, as well as a clear plan for ongoing communications and training.
This layered view of the agentic future is central to an ecosystem built on a strong foundation, carefully orchestrated agent action across the enterprise, and comprehensive governance around APIs, agents, and the whole environment.
Find out more about legacy migration from leaders at Deloitte Consulting LLP and MuleSoft in our on-demand webinar exploring the challenges and opportunities associated with agentic AI transformation.