The Expanding Enterprise Stack, Part 2: The Forecast
Software isn't dying. It's getting a new customer.
This is Part 2 of a series on how AI is reshaping the enterprise. Part 1 laid out the thesis: the stack is growing from two layers (Humans + Software) to three (Humans + Software + Digital Workforces). Digital workforces are net consumers of software, not replacements for it.
This post is about what happens next.
Part 1 landed loud. Mostly agreement. Some pushback. And then someone framed the defensive case better than anyone else has.
The Defensive Case Is Settled
The best counterargument to the “AI kills software” thesis came last week. You cannot replace Salesforce with code a coding assistant generated yesterday. Salesforce has 25 years of bug reports. Maybe millions of them. That system has been tested across thousands of large customers and enterprises. The idea that a small team will rip it out and replace it with probabilistically generated code is not realistic.
The argument extends far beyond Salesforce. Enterprise software isn’t just code. It’s millions of edge cases, resolved. It’s compliance frameworks, battle-tested. It’s integrations with thousands of other systems, maintained across decades.
You can vibe-code a CRM in a weekend. You cannot vibe-code 25 years of enterprise hardening.
But the defensive case only answers half the question. Software isn’t going away. Fine. The more interesting question is the offensive one. What happens next? How do software companies and digital workforce companies co-evolve? And how should enterprises think about investing across both?
Here’s my forecast.
1. How Digital Workforces Will Work With Software
The mental model most people carry is wrong. They imagine digital workforces replacing software. Or replacing humans who use software. Neither captures what’s actually happening.
Digital workforces are a new consumption layer. They sit alongside humans and software, executing work at a velocity humans cannot match. But they depend on software to do it. Every action a digital worker takes requires reading from or writing to a system of record. Every decision requires data. Every output requires a destination.
Think of it this way. A human financial advisor might update a CRM once after a client meeting. A digital workforce processing 5,000 client interactions per month hits that same CRM 5,000 times. Same software. 100x the usage.
We are seeing this at Humanity Labs. One of our wealth management partners had a team of advisors manually processing client service requests. Each request touched their CRM, their portfolio management system, their custodian platform, and their compliance tools. A human might handle 15 of these per day. Our digital workforce handles hundreds. Every single one reads from and writes to the same software stack. The number of API calls to their CRM didn’t decrease when we came in. It multiplied. Their software vendors are getting more usage, not less.
The relationship is symbiotic, not competitive. Digital workforces benefit from three things in software:
Clean data to operate on. The quality of a digital workforce’s output can be improved proportionally to the quality of the data it can access. A CRM with 10 years of interaction history makes the digital workforce dramatically more effective than a blank database. It is not a requirement because no such thing as “clean data” exists, but a spectrum. The more clean the data, the more it accelerates value.
Functions to call. Digital workforces can click buttons AND call APIs. Every software capability exposed as a function becomes part of the digital workforce’s toolkit. The easier these functions are to access, the more symbiotic value will be derived.
Audit trails to write to. In regulated industries especially, every action the digital workforce takes must be logged, attributed, and reviewable. Software systems of record are where that accountability lives.
The pattern is clear. Digital workforces don’t bypass software. They drive more software consumption than humans ever did.
2. What This Means for Software Businesses
Not all software benefits equally. The three-layer stack creates winners and losers within the software category itself.
Winners: Systems of Record
Salesforce, SAP, Workday. These companies sit on decades of proprietary customer data. That data becomes more valuable, not less, when digital workforces need it to operate. The switching costs actually increase because now you’d be disrupting not just human workflows but digital workforce workflows too.
Bank of America made this call last week, upgrading SAP amid the carnage. They’re right. Deep data moats get deeper in a three-layer stack.
Winners: Infrastructure Software
Snowflake, Datadog, Cloudflare. Digital workforces generate enormous amounts of data, require monitoring, and consume compute. More digital workforce activity means more infrastructure load. Full stop. These companies are selling picks and shovels to the new gold rush.
Winners: API-First Platforms
Stripe, Twilio, Plaid. Companies that built their businesses as callable functions are perfectly positioned. They’re already speaking the language digital workforces speak. No translation layer required.
Losers: UI-Dependent Point Solutions
Software where the primary value proposition is “we made a nice interface for X.” If a digital workforce can’t use it easily, and if the underlying function is too simple, it will get absorbed into the orchestration layer.
Losers: Thin Data Moats
Point solutions that don’t accumulate value through usage (eg networks). If the data isn’t getting richer there’s no compounding advantage. The digital workforce doesn’t care which tool it uses for commodity functions, it does care about accessing the best network of tools and systems to accomplish the job, just like a human.
The net effect: the software industry doesn’t shrink. It bifurcates. The strong get stronger. The weak become utilities or disappear.
3. What Each Side Should Invest In
If you’re a software company:
Invest in API coverage. Every feature you have should be callable by a digital workforce. Audit your product surface area. If 60% of your capabilities require a human clicking through a UI, you have a 60% vulnerability. The companies that win the next decade will be the ones where every capability is a function, and the UI is just one of many clients calling those functions.
Invest in data network effects. Make your system the place where data accumulates, compounds, and becomes more valuable over time. This is the real moat. Not the code. Not the UI. The data flywheel.
Invest in digital workforce partnerships. Every major digital workforce provider is looking for software partners with clean APIs, rich data, and reliable infrastructure. Get embedded in their toolkits now. Being the default CRM that digital workforces are trained on is worth more than any feature launch.
Don’t invest in building your own digital workforce. This is the mistake most software companies are making right now. Bolting an “AI agent” onto your product to try to capture both layers. Most don’t have the operational expertise to deliver managed services. Invest in being the best software for digital workforces to operate on.
If you’re a digital workforce company:
Invest in software integration depth. Shallow integrations will commoditize. Deep integrations (understanding the data model, handling edge cases, maintaining state across sessions) are a moat. The digital workforce that can operate a client’s Salesforce instance as fluently as their best employee is the one that wins.
Invest in operational reliability. The point about millions of bug reports applies here too. Digital workforces that operate in production at enterprise scale need the same obsessive focus on reliability that software companies have built over decades. This is not a model quality problem. It’s an operational engineering problem.
Invest in domain expertise. Generic digital workforces are a commodity. A digital workforce that understands the specific compliance requirements, workflow patterns, and data structures of wealth management (or healthcare, or legal, or insurance) is not. Domain specialization is how you build switching costs in a layer that doesn’t naturally have them.
Don’t invest in replacing software. You are better with software than without it. You benefit from its data. You benefit from its APIs. The temptation to “disintermediate” the software layer is real. It’s also a strategic dead end. You would be rebuilding 25 years of enterprise hardening while simultaneously trying to deliver digital workforce value. Pick one.
4. How Enterprise Buyers Should Think About Investment
If you’re an enterprise deciding where to allocate budget, the framework is straightforward.
Double down on your systems of record. The CRM, the ERP, the core platforms where your data lives. These are about to become more valuable, not less. Clean them up. Invest in data quality. Make sure your 10 years of client interaction history is actually usable, not a graveyard of incomplete records.
Audit your software stack for API readiness. Every tool in your stack should be evaluated on one question: can a digital workforce use it easily? If the answer is no, that tool has a shelf life. Start planning the migration now. You don’t need to rip and replace today. But you need a roadmap to an API-first stack.
Budget for digital workforces as a new line item. This is not a software budget. It’s not a headcount budget. It’s a new category. The companies that treat digital workforce spending as a rounding error on their IT budget will lose to the ones that treat it as a strategic capability investment.
Invest in orchestration, not point solutions. Don’t buy seven AI tools for seven workflows. Invest in a digital workforce partner that can deliver across your entire software and business operations stack. The value is in the done work, not the individual automation.
Start with high-volume, low-judgment work. The best use of digital workforces today is work that requires touching many systems, processing many transactions, and following established rules. Not strategic decisions. Not client relationship management. Not exceptions. Volume and velocity first. Judgment and nuance later.
The Punchline
The bears are right about one thing: AI changes everything. But their conclusion is wrong. Software isn’t going away. That’s the boring take. The interesting one is why.
Software isn’t surviving despite AI. It’s becoming more essential because of AI. Digital workforces get better the more software they can operate on. More digital workforce activity means more software consumption, more data generation, more API calls, more infrastructure load.
The enterprise stack is expanding. The companies that understand this will invest accordingly. Software companies will invest in becoming the best platforms for digital workforces to operate on. Digital workforce companies will invest in operating those platforms better than humans can. Enterprise buyers will invest in both.
The ones still debating whether AI replaces software are asking last year’s question. The question now is: how fast does the new layer scale, and who captures the value when it does?