The sexy AI is getting all the attention. The boring AI is getting paid.
Open any tech feed right now and you will see the same five categories of AI content cycling through. The flashy avatars that can speak in any voice. The demo videos of chatbots holding philosophical conversations. The agent demos where someone narrates over their screen while the AI books a flight or writes a poem or composes a song that sounds passable. The image generators producing increasingly elaborate visuals. The voice clones that sound like specific celebrities. These are the AI products that get the press, the social shares, the venture excitement, and the airtime in conference keynotes.
None of them are where the actual money is being made.
Where is the AI money actually being made?
The actual money is being made by the AI that nobody is putting on a stage. It is the AI that reads an invoice and routes it to the right approver, listens to a customer support call and writes the case note, watches a transaction file and flags the missing document before anyone notices, converts a clinical conversation into a structured medical chart, drafts the email follow-up to the procurement team, and reconciles two spreadsheets that used to take a junior analyst four hours to align. None of those examples will trend on social media. All of them are generating real revenue right now.
This is not a hot take. It is the consensus finding of the most rigorous research on enterprise AI adoption to come out in the last 12 months.
What does enterprise AI research show about ROI?
Three of the most rigorous research efforts of the last year point to the same conclusion: the measurable return on investment is concentrated in repetitive workflow categories, not in the flashy consumer-facing AI products that dominate the press. MIT's GenAI Divide, Menlo Ventures' State of Generative AI in the Enterprise, and Andreessen Horowitz's enterprise adoption analysis all converge on the same pattern. The flashy AI is getting the budget; the boring AI is producing the returns.
In August 2025, MIT's NANDA initiative published a report titled The GenAI Divide: State of AI in Business 2025. The research, led by Aditya Challapally, drew on 150 leader interviews, a survey of 350 employees, and analysis of 300 public AI deployments. The report's headline finding was that approximately 95 percent of enterprise generative AI pilots failed to deliver rapid revenue acceleration. Most of the projects stalled with no measurable impact on the P and L. That number got the coverage. The number that mattered more was buried inside the report.
The MIT researchers found that more than half of generative AI budgets across the surveyed organizations were going to sales and marketing tools. The actual return on investment was happening somewhere else. The biggest ROI was concentrated in back-office automation. Eliminating business process outsourcing contracts. Cutting external agency fees. Streamlining the operational workflows that consume human hours every day across every company. The flashy AI was getting the budget. The boring AI was producing the returns.
If you want to know where the real economic shift is happening, this is where you look.
Menlo Ventures published its own State of Generative AI in the Enterprise report for 2025, and the spending data tells the same story from a different angle. Vertical AI became a 3.5 billion dollar category in 2025, triple the dollars invested in the prior year. Enterprise spending on coding tools jumped from 550 million dollars to 4 billion dollars in a single year. IT operations tools reached 700 million dollars as teams automated incident response. Customer success tools captured 630 million dollars, with AI handling ticket routing, sentiment analysis, and proactive outreach. Marketing platforms hit 660 million dollars. The Menlo authors describe the pattern directly. Each of these categories targets repetitive workflows where productivity gains are immediate and measurable.
Read that sentence again. Repetitive workflows where productivity gains are immediate and measurable. That is not the language of demo videos. That is the language of accounting.
Andreessen Horowitz published their own analysis of enterprise AI adoption in April 2026, drawing on internal portfolio data and conversations with corporate executives. They found that 29 percent of the Fortune 500 and approximately 19 percent of the Global 2000 are now paying customers of a leading AI startup. The categories where the revenue is concentrated are exactly the categories the MIT and Menlo data points to. Coding, where AI accelerates engineer productivity. Legal, where AI handles dense unstructured text. Healthcare, where AI scribes turn clinical conversations into structured medical records. Customer support. Operations. The boring categories. The categories where the work is repetitive, the workflow is mappable, and the value is measurable.
The a16z report also noted something I want to highlight specifically. They wrote that AI is excellent at parsing dense text, reasoning over large amounts of text, and summarizing and drafting responses. All work that lawyers regularly do. The same logic applies to accountants, transaction coordinators, paralegals, claims adjusters, agents, brokers, compliance officers, and every other professional category whose day is structured around document-heavy, language-heavy, repetitive operational work.
The boring AI is going after exactly those workflows. The flashy AI is going after attention.
Why does most of the industry get this wrong?
Most of the industry pictures AI making money through novelty. Futuristic product lines, AI talking heads, agents that book travel, experience-layer products that sit between the human and the world. That picture is wrong. The money is in the workflow layer underneath the experience, inside the operations of every business that runs on documents, decisions, communications, and structured tasks.
The AI that takes an hour out of a transaction coordinator's day. The AI that takes a day out of a closing process. The AI that takes a week out of a quarterly close. The AI that takes the friction out of the thousands of small, language-heavy, document-heavy operations that every real business runs on. Nobody films a demo video about that AI. Nobody posts a reaction reel about it. Nobody tags their friends to come look at what it did. It is too boring to share.
It is also where every category-defining business of the next decade is going to make its money.
The reason is structural. Sales and marketing AI has to compete with humans on creativity, where humans still have an edge in many cases, and the value is hard to attribute cleanly to the tool. Was the increase in click-through rate from the AI or from the campaign timing? Hard to say. Back-office AI does not have that problem. Back-office AI shows up as cost reduction, throughput increase, error rate decrease, or compliance improvement. The math is clean. The math is verifiable. The math goes straight to the P and L. Operators can see it on their dashboards and CFOs can see it on their reports.
This pattern shows up everywhere the data has been studied carefully. It is showing up in finance, where AI is replacing the layer of work that used to be done by junior analysts reconciling data across systems. It is showing up in healthcare, where AI scribes are saving physicians ninety minutes a day on documentation. It is showing up in legal, where AI is summarizing case law and drafting contract redlines for review. It is showing up in customer support, where AI is triaging tickets and writing first-draft responses. It is showing up in IT, where AI is detecting incidents and suggesting fixes. It is showing up in operations, where AI is doing the reconciliation and routing work that used to require teams of coordinators.
How will boring AI compress real estate work?
Real estate transactions are structured around exactly the kind of document-heavy, language-heavy, repetitive workflow that boring AI is built to compress. Document review, form completion, communication drafting, compliance flagging, disclosure handling, calendar reconciliation, and multi-party status updates make up the majority of an agent's working hours. The agents who deploy operational AI inside these workflows over the next 24 months are the ones who will pull ahead.
None of those tasks make for a flashy demo video. All of them are where agents spend the majority of their working hours, and all of them are now compressible by an order of magnitude.
The agents who pull ahead in the next 24 months will not be the ones who buy the flashiest AI marketing product on the market. They will be the ones who deploy the boring AI inside the operational workflows that consume their day. The agents who watch the flashy demos and wait for the perfect consumer-facing tool will keep doing the same work the same way, and they will fall further behind every quarter as the operator class compounds capability through the unglamorous categories, the same way practice compounds capability for any operator who keeps showing up to the work.
I want to be precise about something here. I am not saying the flashy AI is fake or useless. The capabilities are real. The breakthroughs are real. The flashy AI is impressive and will continue to be impressive. The flashy AI is also where most of the airtime, marketing budget, and consumer attention will continue to flow. None of that is wrong.
What I am saying is that if you are an operator, a CEO, a broker, an agent, or a professional in any industry that runs on routine knowledge work, your money question is the boring AI question. Where in your workflow is the next hour, the next day, the next week of human time about to get compressed by a system that nobody will post about on LinkedIn? That is the question that will define your competitive position over the next decade. Not which avatar can speak in your voice. Not which agent can book your dinner reservation. The boring questions are the expensive questions, and the boring answers are the ones that are about to determine which businesses win and which businesses get left behind.
The future is great with the flashy AI. The money is in the boring.
That is what every conversation about AI in business right now should actually be about.
Judd Hoffman is CEO and Co-Founder of Ethica AI, building AI-powered tools for real estate transaction workflows.