Most operators already have AI on every laptop. Here’s what they’ll pay for on top of it in 2027, based on real questions from OG Demo Day.
In a recent CFO Connect survey of 215 finance leaders, Claude was the AI tool finance teams cited most, at 41%. No specialist finance AI tool cleared 1%.
That’s the backdrop for 2027 planning. Most operators already have a capable AI on every laptop, and some are building their own tools with it. So the question in every vendor conversation is now “what’s worth paying for on top of what we already have?”
We saw that play out at OG Demo Day in September, where operators watched vendors demo across ERP, FP&A, the close, billing, people ops, and AI agents. The questions they typed into the Q&A box made their priorities pretty clear. Here are five priorities shaping operator tech stacks going into 2027.
1. A clear answer to build vs. buy
With frontier models a login away, every specialized tool now competes with what a team could build itself. NeuGroup reports treasury teams building working tools on general-purpose AI like Claude without waiting on a vendor or IT. Buying still wins most of the time in finance, though. The Hackett Group found embedded vendor AI outnumbers custom builds about 6 to 1 in accounts payable.
At Demo Day, this question followed vendors into every session:
- “My team has been experimenting with Claude. How is this different?”
- “Why couldn’t the agents already in my ERP do this?”
- “When is Claude or Codex good enough, and when do I need a separate product?”
- “What’s the difference between your routines and the ones I can build in Claude Code?”
- “Is everything in the app also available through MCP?”
Operators want to know what a specialized tool takes off their plate. As one Forbes Finance Council piece put it, the real question is “Do we want to own everything required to trust it over time?” The vendors who answered well pointed to finance-specific context, built-in controls, and maintenance a lean team doesn’t want to take on.
2. Human in the loop, by design
Across the FP&A, close, and billing sessions, one question kept coming up: who stays in control? That matches Deloitte’s Finance Trends 2027, which found that 95% of finance leaders are comfortable with agentic AI in at least some activities, but only 14% support full autonomy for critical decisions. And 42% say their ambitions are ahead of their capabilities, citing the need for stronger controls first.
Operators asked:
- “If the AI misreads a contract clause, how do we catch it before a wrong invoice goes out, and is there a record of what it changed?”
- “Is there any risk of made-up numbers when I ask the AI for figures?”
- “My team is excited about agents, but how do I make sure my AP analyst can’t search payroll data?”
- “Do agents start with permissions closed until someone opts in?”
- “Can a human always override the agent’s recommendation?”
For 2027, controls are part of the product, including review steps, change logs, permissions that start closed, and deterministic math underneath the AI. The strongest answers were specific, like calculations running in a deterministic system and audit trails built to hold up to SOX review.
3. Tools that fit the stack you already have
This is a consolidation year. Teams are cutting overlapping apps. Some are finding their total bill hasn’t moved, The Spend Ledger notes, because usage-based AI pricing on the tools they keep expands to fill the gap. A new tool has to slot into what’s already there.
At Demo Day, fit with the existing stack came up in every category:
- “How painful is it actually going to be to get our data out of QuickBooks?”
- “Can we run the new system in parallel for a cycle before cutting over?”
- “How quickly could a NetSuite team be running this on its own data?”
- “Does this replace our CRM or sit alongside it? Does it work with Ramp, Brex, Salesforce, and our bank?”
- “We have a report we build in Excel that’s specific to us. Can your team help recreate it?”
This was loudest in the ERP session, where switching carries the most risk. Operators are protecting what works today, like payroll, tax filings, and the month-end close, and they want tools that connect to it cleanly.
4. Specialists that handle the hard cases
General-purpose AI is good at the common path. The exceptions are where operators expect a specialist to earn its keep, and they tested vendors on exactly those:
- “We bill some customers in USD and others in EUR under one contract. How does revenue recognition handle FX without manual journal entries?”
- “If a customer upgrades mid-contract and gets a one-time credit, does the schedule update automatically, or does someone rebuild it?”
- “We’re expanding into Europe and India. How do you handle localization and e-invoicing rules?”
- “How does this work across entities with different accounting policies?”
The hard cases are also multiplying. Several attendees said their own companies are moving to AI credits and usage-based pricing and struggling to bill for it accurately. Simon-Kucher describes how one credit becomes three numbers for finance: committed, consumed, and recognized. Deloitte found 60% of finance leaders expect to need more sophisticated AI cost management through 2027. AI is changing what these companies sell and how they charge for it, and that lands on the finance team’s desk.
5. A clear path to adoption
Getting a team to use an AI tool well is where a lot of operators are stuck, and it’s one of the topics we hear about most across the OG community. Finance increasingly owns that plan. Deloitte found 54% of finance leaders now lead cross-enterprise AI and technology capital allocation, and among leaders whose AI ambitions are ahead of their readiness, 41% name employee trust in the quality of AI outputs as a leading barrier.
At Demo Day, the questions about AI agents were about rollout more than features:
- “Is this something I’d deploy for my whole org, or something one operator sets up for themselves?”
- “Do I need to clean up my CRM before any of this works?”
- “What data and integrations are necessary, and which are nice to have?”
- “What quantifiable results have customers seen?”
- “How is privacy handled once it’s connected to email and calendar?”
Operators want to know who owns it, what has to be in place first, how you prove it worked, and whether people trust it enough to use it. For 2027, they want tools that come with an adoption path: clear prerequisites, a way to start with one team and expand, and results they can measure.
The 2027 stack
Operators are buying, and they’re buying differently: a general-purpose AI the whole team uses, plus a smaller set of specialists that earn their place on trust, fit, the hard cases, and adoption.
These are the questions operators are working through for 2027, and that’s a big part of why Demo Day exists. Operators can see a whole category back to back, alongside peers asking what they were about to ask.
OG members can watch every Demo Day session in their resources space. Not a member yet? Apply here.