Three years into the AI boom, the companies drawing the most investor attention are still the ones that are easy to see. Chatbots answer questions. Agents complete tasks. The visible layer of the AI economy has driven most of the narrative.
But a different kind of AI is gaining ground, and it mostly goes unnoticed. It sits inside the systems that generate invoices, move money between organizations, assess credit exposure, and capture billable time.
Nobody calls it AI when it’s working. They just notice that fewer bills get rejected and collections take less time.
That gap between where AI gets attention and where it may actually be creating value is increasingly relevant for investors. Legal services is one place where the gap is narrowing fast, and the companies enabling it are starting to look like a serious infrastructure play.
Also read: Apple’s biggest AI advantage may face new privacy challenge
The AI that works best is the kind no one sees using
Most enterprise software still requires people to adapt to it.
The more interesting shift happening now is the reverse: Intelligence is being embedded into workflows so deeply that the people inside those workflows never have to think about it.
A lawyer doesn’t need to know that AI reviewed an invoice against a client’s billing guidelines before it went out. A finance director doesn’t need to know why exceptions are down. The value shows up in the outcome, not the interface.
Ahmed Shaaban, CEO of legal operations technology company Fulcrum GT, argued that this is where enterprise AI becomes economically meaningful.
“We’ve shown the world that AI can hold a conversation,” Shaaban told TheStreet in an interview. “Now let’s put it to work. Connect it to how a business bills, manages risk, and gets paid, and you start opening up possibilities that weren’t practical before. That’s the next chapter: intelligence working inside the business, helping people do things they simply couldn’t do before.”
The challenge is that embedding intelligence into operational processes requires the underlying data and systems to already be in reasonable shape. If billing, time capture, and matter management live in disconnected silos, adding AI exposes those gaps rather than closing them.
Law firms are building the operational foundation first
Some of the most established law firms in the world are already several years into that foundational work. Their experiences are instructive not because they represent the finished product, but because they show how much has to come before any AI deployment can actually deliver.
DLA Piper International began with process standardization and a shared services center in Warsaw. The firm established clear senior ownership over its finance function, implemented enterprise technology, and built a consistent internal team capable of carrying transformation across borders.
Garry Swaine, who led the effort, said the platform went live across 30 countries on the same day. The foundation made the speed possible.
Eversheds Sutherland took a similar route, consolidating around core platforms that could grow alongside the business. The firm spent 180 hours consulting directly with lawyers about the friction they experienced in daily work.
That exercise surfaced a collections process requiring roughly 300 records to be updated by hand each month, a problem the firm is now exploring whether an AI agent could handle without overhauling the broader collections model.
Both firms described their AI applications as areas still under development. The platforms and processes are in place. The efficiency gains from layering AI onto them are what comes next.
A new ownership model is entering the picture
Alongside traditional partnerships investing through their own structures, management services organizations (MSO) are emerging as a separate path to modernized legal operations.
An MSO provides shared non-legal services, including finance, technology, and administration, through a separately organized entity, while leaving the law firm itself in the hands of its lawyers.
More AI:
- Nvidia just made a move Wall Street wasn’t ready for
- Microsoft just took sides in AI policy fight
- OpenAI just disclosed something genuinely alarming
Orion Legal MSO, formed in January 2026 by Uplift Investors, offers one early example. Louisiana personal injury firm Dudley DeBosier joined as the founding partner.
Uplift has said the structure is designed to deliver operational support without altering firm ownership or control. The stated investment thesis includes technology, data, and AI integration across the firms the MSO supports.
Whether the MSO model outperforms a well-run partnership investing in its own shared services is not yet clear.
The DLA Piper and Eversheds examples show that established firms are capable of significant operational transformation without outside capital or a separate organizational structure. The MSO model adds a path; it does not establish a superior one.
What enterprise software investors are paying attention to
For investors assessing enterprise software companies, the dynamics playing out in legal services reflect a broader question about where AI creates durable commercial value.
An established firm expanding internationally, a partnership improving its existing operations, and an MSO building shared infrastructure across multiple firms may all need the same underlying capabilities. Demand can come from several directions at once.
Scott Mozarsky, a former president of Bloomberg Law, frames the investment question around proximity to mission-critical workflows. “From an investment perspective, simply being able to say that a product contains AI is becoming less interesting,” Mozarsky told TheStreet in an interview.
“The more important questions are whether the company sits inside a mission-critical workflow, whether it has access to differentiated data and context, and whether its technology is connected closely enough to the customer’s operations to create measurable economic value.”
That framing is useful precisely because it separates the hype from the signal. A software provider’s position in a critical process is a starting point, not a verdict.
Investors still need to examine how deeply embedded the product actually is, what it costs to implement, whether adoption is genuinely sticky, and whether the efficiency improvements being promised translate into numbers the customer can point to.
Governance cannot be added after the fact
The deeper AI goes into operational processes, the higher the stakes become when something goes wrong. A system that can change a financial record or initiate a payment needs more than a capable model. It needs reliable data, defined access controls, and a clear path for human intervention before errors compound.
At Eversheds Sutherland, compliance, cybersecurity, and in-house counsel were part of the delivery team from the start.
That is a practical acknowledgment that the governance layer cannot be retrofitted once AI is already embedded in a workflow. It has to be part of the design from day one.
That requirement raises the bar for any software provider looking to move AI deeper into legal and financial operations. The firms that make it work will likely be the ones that treated data quality, permissioning, and auditability as engineering problems rather than policy ones.
The firms that try to add controls later will find out why that order of operations matters.
