Generative artificial intelligence will increase the risks of disruption anywhere data is reasonably available, and where value is extracted by organizing or analyzing the internal logic of that data, according to Morningstar equity research.

Coding is the most obvious example of this. Though the same could apply to certain aspects of legal analysis or medical research, humans are trying to organize logical connections using mostly publicly available text.

Morningstar equity research recently reevaluated the moat ratings for 132 companies across our coverage, where we felt AI could be disruptive and therefore warranted deeper analysis.

Our evaluation included a look at arguments we hear in the great gen AI versus software-as-a-service debate: the idea that AI will replace traditional software solutions. One of our own key takeaways is that no one knows for sure.

There are good points on both sides. Here’s a look at the seven points and counterpoints in the gen AI versus SaaS debate.

AI lowers the cost of producing code

AI is really good at coding and consequently lowers the economic surplus that companies can derive from coding. Whatever percentage of the value that a software company adds from pure coding, then less of that value will accrue to the software providers once coding becomes less of a scarce resource.

Counterpoint

The value of SaaS is not just from the coding. The value is from complex things that AI is not as good at, like knowing the business context, domain expertise on specific workflows, accounting for numerous edge cases, general ownership and monitoring, and more. The database schema, history, security, compliance, and trust are what matter.

At the end of the day, companies pay for SaaS because “it just works,” and getting the code is only one piece of that puzzle.

Software is often built on ‘per seat’ economics, which is directly threatened

AI should increase the productivity of the end user or even displace the end user. This should lower the seat counts needed over time. If one person can do the work of two people, that could lead to 50% fewer seats.

AI could also lower the price of whatever is being sold. If the price of “work” is reduced to $20 a month per user, and your current pricing is more expensive than that, there could be price compression.

Counterpoint

Many SaaS companies already charge based on “value” or “usage.” Those that don’t will transition their pricing models in an agentic world. If a company attempts, it could still face price compression if it was previously earning economic rents disproportionate to the value added that AI offers, but it will fend off any of the more dire scenarios.

Additional demand isn’t out of the question. Historically, an increase in worker productivity has not led to a severe contraction in the workforce. Instead, companies use the same number of workers, and they are simply able to do more.

Gen AI is based on token consumption and has lower gross margins

AI is fundamentally a consumption-based model. You are consuming intelligence via consuming compute. Software’s main strength was essentially no incremental cost for distribution. That is gone in a consumption-based world. The higher the percentage of your revenue stream that comes from consumption-based AI, the more your gross-margin profile will drift toward consumption-based gross margins.

Software gross margins could go from 80% or more to 65% or less.

Counterpoint

The cost of AI keeps going down as new architectures and better hardware emerge. The value produced by each new model keeps rising. The cost of improving inference will still allow for good gross margins, and companies will use model routers to apply appropriately priced tokens to relevant workflows.

AI products will expand the total addressable market, which will still increase the absolute dollar amounts that software firms can address. Finally, software firms can also increase their efficiency using AI, so any gross margin pressure could be offset by operating efficiencies, leaving operating margins intact.

Gen AI lowers barriers to entry

If coding is mostly automated, then the barriers to creating things with code become much lower. This means it is more likely that SaaS clients will create internal solutions to replace existing software. Also, any company with software engineers will start creating products for additional verticals, which will increase competition.

Lower barriers to entry will lead to a larger number of competitors. Not only that, but these competitors will also accelerate their own product roadmaps, as AI improves product velocity.

Counterpoint

The true barriers to entry aren’t just “producing the code.” It’s all the other things that SaaS companies do—providing maintenance, security, complex coordination and implementations—that shouldn’t be discounted. Owning existing client relationships and having a competent go-to-market process still matter to get clients to adopt a company’s offerings.

AI lowers switching costs

By making processes easier and automated, AI lowers switching costs wherever there are difficulties in changing software products due to technical hurdles, like remapping data.

Counterpoint

While AI arguably should be able to automate some aspects of data transfer, you still need the internal expertise to prompt and manage the AI when doing even moderately complex data migrations.

AI changes the basis of competition and industry architecture

Key bases of competition for software are the talent of a company’s developers, product-market fit for clients’ needs, and executing a sales strategy. If AI becomes central to future developer platforms, AI competency becomes a key basis of competition.

The industry architecture could also change. If AI becomes the central platform through which workflows run, it has the potential to remap current winners and losers.

As AI models become central, more value will accrue to the model providers than before. (The value accrued to model providers was previously $0.) Model providers can now “one-shot” different workflows as individual integrations within their model. As they integrate more workflows, the model only becomes more powerful.

Counterpoint

Once again, software is about much more than “the code.” Success depends on knowing the business context, having domain expertise on specific workflows, and actually selling and distributing your product.

Technology shifts introduce uncertainty into the system.

At a minimum, AI developments are introducing a lot of change. Today’s software companies will have to adapt. When a system changes, it’s hard to predict who figures it out and who doesn’t. There are many reasons why incumbents have a harder time adjusting to changing paradigms, including technological debt and corporate inertia.

Uncertainty has risen for everyone. This feels like trying to predict the SaaS winners in 2001—it wasn’t easy.

Counterpoint

There is no fundamental reason that incumbents can’t adapt and benefit from AI. Incumbents could turn AI services into a tailwind, not a headwind, as they adopt AI and turn them into new features that generate revenue, maintaining their place in the ecosystem.