The following is a guest post from Neima Golnabi, partner, finance transformation at Grant Thornton. Opinions are the author’s own.
Every CFO who has been through an acquisition knows the moment. The deal is done and the integration plan calls for consolidating onto a single ERP. Then the price tag lands on your desk: tens of millions of dollars, two or three years of work, significant disruption to the business and an ROI that's difficult to defend to a board looking for results this year.
So the second and third ERPs stay. The close gets a little slower. Reconciliations become a little more manual. Reporting requires more mapping and workarounds. And now there's another question coming from the board: What are we doing with artificial intelligence?
Here's the challenge. Fragmented systems and inconsistent data are among the biggest barriers standing between finance teams and meaningful value from AI. But that doesn't mean you need to launch a massive ERP consolidation before you can get started.
There is a more practical path: sequence the work so AI creates value along the way and helps build the business case for the broader transformation.
The ambition is there; the foundation often isn't
Finance leaders are clearly investing. In Grant Thornton's Q1 2026 CFO survey, 68% of finance leaders said they expect IT and digital transformation spending to increase over the next year, the highest level in the survey's 21-quarter history. Nearly all organizations, 97%, report that they are piloting, scaling or fully integrating AI.
But the payoff hasn't always kept pace with the investment. Only a third of finance leaders in Grant Thornton's September 2026 survey reported enhanced revenue as an AI benefit. In a Coupa survey, 76% of CFOs cited unclear ROI as a considerable barrier to further AI investment.
Why the disconnect? One reason shows up repeatedly: poor data quality and fragmented systems. Datarails' 2026 research found that nearly a quarter of finance teams still rely on disconnected systems and manual reconciliation. AI doesn't magically fix messy data. In many cases, it exposes the problem faster. An AI agent working across three charts of accounts, three vendor masters and three sets of cost-center definitions doesn't automatically create one version of the truth. Without a common data foundation, it may simply produce three answers faster.
The hidden cost of waiting to consolidate
Most CFOs already understand the operational cost of multiple ERPs. Complex, multi-ERP environments can make that challenge even harder: manual intercompany eliminations, offline mapping files, inconsistent master data, and reporting teams spending too much time reconciling numbers instead of explaining what they mean.
There is also a control consideration. System transitions and partial integrations can create gaps when controls designed for one environment don't translate cleanly to another, but AI introduces another cost that is harder to see.
Every year, data remains fragmented, making it harder to move AI beyond isolated pilots and embed it into core finance processes. Consider a fast-growing, private equity-backed manufacturer operating multiple ERPs. The company may understand the benefits of consolidation but simply not have the capacity to take it on immediately. Each ERP conversion could take 18 months or more because of process differences, data complexity and the change management required to make the transition successful.
A better sequence: data first, consolidation second
Waiting until every ERP is consolidated before pursuing AI could mean waiting years, but there is another option. The all-or-nothing framing is often what makes these programs difficult to get off the ground. A more pragmatic approach can happen in three stages.
1. Standardize the data, not necessarily the systems. Start by defining the common language of finance: a standard chart of accounts, vendor and customer master-data standards, and consistent entity, product and cost-center definitions.
Then map the existing ERPs to that common structure through a shared data layer. This doesn't eliminate the underlying systems, but it can give finance a more consistent foundation for consolidation, analytics, reporting, and, eventually, AI.
2. Put AI to work where the data is ready. Once the data speaks a more common language, AI can begin operating across systems without requiring every ERP to be replaced first.
Accounts payable invoice processing, account reconciliations, intercompany matching, transaction classification, anomaly detection and variance commentary are good places to start. They tend to involve large datasets, repetitive activities and identifiable exceptions, making the results easier to measure. This is where the opportunity becomes particularly compelling.
The greater the transaction volume and operational complexity, the stronger the potential AI use case. In multi-entity, multi-jurisdiction and multi-currency environments, finance teams are constantly processing transactions, applying rules, reconciling balances and looking for exceptions.
AI can help continuously analyze those datasets, identify unusual transactions and patterns, automate repetitive decisions and, importantly, direct people toward the transactions that actually require human judgment. The objective isn't to remove finance professionals from the process. It's to stop asking them to spend valuable time finding the exception before they can begin solving it.
3. Use early wins to help fund the broader transformation. This is where the economics start to change. If AI can reduce reconciliation effort, shorten portions of the close, lower transaction-processing costs or create capacity within the finance organization, those benefits can be measured. Now the CFO has something much stronger to take to the board: demonstrated value rather than projected value.
ERP consolidation can then happen in phases, prioritizing the systems and processes where the business case is strongest instead of approaching transformation as one large, multiyear bet.
Governance can't be an afterthought
Governance and integration aren't barriers to AI value. They are part of what makes sustainable AI value possible. For finance organizations, that means understanding where the data came from, who owns it, what rules are being applied, how exceptions are handled and where human review remains necessary. The goal isn't simply to deploy AI. It's to deploy AI that finance leaders, auditors and the business can trust.
If your organization is operating more than one ERP, ask three questions before approving the next AI pilot:
- Do we have a common chart of accounts and master-data standards across our systems?
- Which high-volume finance processes could benefit from AI today if the underlying data were consistent?
- Can we measure those benefits clearly enough to help fund the next phase of transformation?
The CFOs who get the most value from AI over the next few years won't necessarily be the ones who spend the most on new technology. They will be the ones who create a trusted data foundation, apply AI where it can solve real finance problems, prove the value and use those results to fund what comes next. You don't have to wait for a perfect ERP landscape to start, but you do need to give AI a consistent version of the truth.