What boring SaaS will people always need?

Why is it suddenly harder to know which boring SaaS will survive?

Boring SaaS is much harder to judge today because AI can wipe out a software product without wiping out the job that product was built around.

That distinction changes almost everything. Businesses will keep reconciling accounts, scheduling workers, creating invoices, approving bills, signing contracts and filling forms. But some of those activities no longer justify a separate subscription.

We can see the shift inside the incumbents themselves. QuickBooks is putting AI into accounting work that once required users to categorize transactions and investigate discrepancies manually. ServiceTitan is adding AI across calls, customer histories, estimates and field-service workflows. Procore is building agents around RFIs, submittals, contracts and construction records.

Those companies are automating their own interfaces because the database underneath them is more valuable than the clicks.

That gives us a tougher test for boring SaaS. A recurring task is no longer enough. We want a recurring obligation where somebody still needs an authoritative record after the AI has finished doing the work.

Payroll taxes still have to reconcile. Financial statements still need a ledger behind them. A contractor still needs to know what happened at a customer's house. A pharmaceutical company still needs controlled records. An employer still needs to know who has access to payroll data.

The apps around those obligations can change dramatically. The underlying systems are much harder to remove.

What does “people will always need it” actually mean for SaaS?

An “always-needed” SaaS category has a business obligation that keeps coming back regardless of which interface, vendor or AI model handles it.

We should be careful with the word “always.” ADP, QuickBooks or ServiceTitan themselves are obviously not guaranteed to exist forever. What matters is whether another system would have to replace them if they disappeared.

Payroll passes that test. Employers continue generating wages, deductions, tax liabilities and employment records every pay cycle. In the United States, most employers file federal employment-tax returns every quarter, while wage and hour records have statutory retention requirements.

Accounting passes too. Companies can automate bookkeeping almost completely and still need an agreed financial record of what they own, owe, earned and spent.

Basic appointment scheduling is different. The activity survives, but the scheduling product can be swallowed by a field-service platform, payroll suite, healthcare system, CRM or AI assistant with almost no loss of information.

The useful distinction is whether removing the product destroys authoritative business history or merely removes a convenient way of interacting with it.

SaaS job	How durable is the underlying need?	Risk that the standalone app disappears
Payroll and employment tax	Extremely high	Low
Accounting ledger and close	Extremely high	Low
AP, AR and payment controls	Very high	Low to medium
Regulated records and compliance	Extremely high	Low
Vertical operating systems	Very high	Low
Security and access control	Very high	Medium
Contract management	Very high	Medium
Scheduling alone	High	High
Basic invoicing alone	High	High
Generic forms and reporting	High	Very high

Is payroll the safest boring SaaS business today?

Payroll is probably the safest boring SaaS category because employers can change payroll providers, but they cannot decide that paying workers and reporting wages is optional.

The latest numbers remain unusually strong for such an old software category. ADP finished its latest fiscal year with $21.9 billion of revenue. Employer Services client revenue retention was 92.1%, and ADP says retention at that level implies an average Employer Services relationship of roughly 13 years.

Paycom shows the same pattern farther down the market. Its latest annual filing reports roughly 39,200 clients and 91% annual revenue retention, up from 90% the previous year. Those clients held records for more than 7.4 million employees.

Paychex adds an important complication. It currently serves roughly 840,000 customers, including about 800,000 payroll clients, yet payroll-client retention runs around 82% to 83%. That looks much weaker until we consider who the customers are. Smaller businesses close, merge or stop employing people far more frequently than established mid-market companies. U.S. Bureau of Labor Statistics data has shown that only around a third of establishments survive ten years.

So even one of the most unavoidable SaaS problems does not guarantee perfect customer retention. Customer mortality still matters.

What makes payroll unusually durable is the combination of recurring calculation, money movement, taxes, filings, employee records and historical corrections. AI can automate huge parts of payroll administration and probably will. Somebody still has to own the final payroll record.

The current evidence also suggests that customers increasingly buy more around payroll rather than replacing it. Paychex now supports about 2.6 million worksite employees through HR outsourcing and retirement plans covering roughly 1.6 million employees. ADP continues expanding from payroll into global HR, time and workforce products.

Payroll increasingly acts as the spine for a broader employment system.

Will businesses still pay for accounting SaaS when AI does the bookkeeping?

Yes. Accounting SaaS looks safer today than many AI-disruption arguments suggest because companies need a reliable ledger even if machines perform most of the bookkeeping.

Intuit's newest annual filing gives us a useful real-world test. QuickBooks Online Accounting revenue reached about $5.05 billion, up 23% in one year. Two years earlier it was roughly $3.38 billion.

That means QuickBooks Online Accounting has added around $1.67 billion of annual revenue in two years, or almost 50% of its starting base, during the same period in which generative AI became far better at transaction classification, extraction, reconciliation and financial Q&A.

Intuit is also building those capabilities directly into QuickBooks. Accounting agents can help categorize activity, reconcile accounts, flag unusual transactions and chase missing information. The manual bookkeeping layer is already being attacked from inside the product.

Yet the accounting business continues to grow quickly.

The reason becomes clearer when we separate bookkeeping from accounting infrastructure. An AI can decide that a card charge looks like software expense. A company still needs that decision recorded consistently in a ledger connected to its bank transactions, invoices, payroll, taxes, reporting periods and previous entries.

The latest Intuit numbers make the broader platform effect visible too. Its Online Ecosystem generated about $9.9 billion in the latest fiscal year, versus $6.9 billion two years earlier. QuickBooks has expanded well beyond ledger software into payroll, payments, financing and other financial services.

AI appears to be taking work away from bookkeepers faster than it takes value away from the accounting system itself.

Will businesses always pay for accounts payable and accounts receivable software?

Businesses will keep paying for AP, AR and payment-control software because paying suppliers and collecting customer money are unavoidable, recurring events with real financial consequences.

BILL gives us one of the clearest measures of how much economic activity can sit behind this kind of boring software. Its latest annual filing says roughly 479,300 businesses used its solutions and processed $371.3 billion of payment volume over the fiscal year. About 9.2 million network members had electronically paid or received money through the platform.

That is roughly a billion dollars of payment volume per day averaged across the year.

The valuable part goes well beyond generating an invoice or displaying a “Pay” button. BILL sits between invoices, approvals, vendor records, accounting systems, bank accounts, cards, payment processors and expense policies. Removing that control layer can mean rebuilding permissions, payment instructions, approval chains and accounting integrations.

There is still real bundling risk here. BILL itself is pushing embedded capabilities through platforms including Paychex, Oracle NetSuite and Acumatica. QuickBooks keeps adding money products. Banks and vertical SaaS companies are moving in the same direction.

So we would be much more confident building around payment control than around invoice creation.

The invoice survives. A standalone invoice generator has a far shakier future.

Is compliance SaaS one of the safest businesses against AI?

Compliance SaaS can be extremely durable when the software proves what a company actually did, while products that merely generate compliance documents look much weaker.

The difference is evidence.

Financial institutions covered by the FTC Safeguards Rule, for example, can face requirements around written security programs, risk assessments, access controls, system monitoring, incident-response plans and continuing reassessment. GDPR similarly forces organizations handling personal data to demonstrate accountability rather than simply claim that they are compliant.

Enterprise customers create another layer of pressure. Large buyers routinely ask vendors for SOC 2 reports, ISO certifications, security questionnaires, access-control evidence and policy documentation before purchasing software.

Recent Vanta customer examples show how commercial that requirement can become. Snowfire said seven sales opportunities had stalled while it lacked the necessary SOC 2 evidence, including a multimillion-dollar opportunity. Hyperbound said ISO 27001 compliance helped unlock a $1.5 million customer and opened conversations with much larger enterprises.

Vendor case studies are examples rather than market averages, but the mechanism is straightforward. Compliance can sit directly in the path between a company and revenue.

AI will probably erase a lot of compliance busywork. It can already draft policies, answer questionnaires, identify missing evidence and map controls across frameworks.

That makes a generic “AI compliance writer” easy to copy. The stronger product continuously knows which employee has which access, which cloud system is configured how, which controls passed, which failed, what evidence exists and when the evidence was collected.

Auditors and customers ultimately care about that history.

Why is vertical SaaS holding up better than generic SaaS?

Vertical SaaS is holding up well because the best platforms know how a specific business actually operates, including details that a general AI model cannot infer from the internet.

ServiceTitan is a particularly clean example. Its latest annual filing says gross dollar retention has stayed above 95% for three consecutive fiscal years. The company now has roughly 10,800 active customers, up from about 9,500 a year earlier, while gross transaction volume handled through the platform reached $82.1 billion over the latest fiscal year.

Those customers are not buying a prettier calendar. HVAC, plumbing and electrical contractors use ServiceTitan across customer calls, scheduling, dispatch, technician workflows, estimates, invoices, payments, memberships and reporting.

A general AI model may know how an air conditioner works. It does not automatically know which system is installed at 18 Oak Street, which technician serviced it two years ago, what was replaced, whether the customer has a maintenance plan, what price was quoted yesterday and whether payment arrived.

Procore shows the same dynamic in construction. Its most recent quarterly filing reports 95% gross retention. The number of customers paying more than $100,000 of annual recurring revenue has climbed from 2,517 to 2,871 in one year, an increase of 14%.

That growth is happening while Procore adds AI across project records, RFIs, contracts and other construction workflows.

When AI enters a vertical SaaS product, it needs the customer's operating data to become genuinely useful.

Owning that data puts the incumbent in a much stronger position than owning a generic interface.

Which boring vertical SaaS markets look strongest right now?

Trades, construction, restaurants, fleets and life sciences currently stand out because their software touches physical work, money or regulated records several times a day.

Toast is a good restaurant example. Its latest quarterly results show roughly 180,000 locations, up 22% in a year. Those locations processed $215 billion of gross payment volume over the previous twelve months. Annual recurring run-rate reached about $2.4 billion, up 25%.

Restaurants still have to accept orders, route them to kitchens, take payments, manage menus, schedule staff and reconcile sales. AI can automate individual decisions inside those workflows, but it does not remove the operating system connecting them.

Samsara shows a similar effect in fleets and industrial operations. The company ended its latest fiscal year at $1.9 billion of annual recurring revenue, up 30%, with 3,194 customers spending more than $100,000 each per year. Those large accounts alone represented about $1.2 billion of ARR.

The interesting part is how broad adoption becomes once a physical-operations platform gets established. Samsara says 96% of its $100,000-plus customers use at least two products, and 69% use at least three. Vehicles, cameras, drivers, maintenance, safety and routing naturally pull the software into adjacent workflows.

Procore provides another version of the same effect. By the end of its latest full year, 78% of ARR came from customers using at least four Procore products, while 52% came from customers using six or more.

That is more convincing than a generic SaaS company simply adding modules. The customer is gradually concentrating operational history inside one place.

Vertical	Recent evidence	Why the software stays useful
Trades	ServiceTitan >95% gross dollar retention; $82.1B annual GTV	Jobs, customers, technicians, estimates and payments share one history
Construction	Procore 95% gross retention; 2,871 customers above $100K ARR	Projects create years of documents, approvals and contractual records
Restaurants	Toast ~180,000 locations; $215B trailing GPV	Orders, POS, staff and payments happen continuously
Fleets	Samsara $1.9B ARR; 3,194 customers above $100K ARR	Physical assets continuously generate operational data
Life sciences	Veeva $2.68B subscription revenue	Regulated records and processes persist for years

Is life-sciences SaaS even harder to replace than ordinary enterprise software?

Life-sciences SaaS is one of the hardest categories to displace because pharmaceutical companies need controlled records long after the employee who created them has moved on.

Veeva gives us unusually strong evidence. Subscription revenue grew from roughly $1.90 billion to $2.28 billion and then $2.68 billion over the past two fiscal years. That works out to about 41% cumulative growth from the first figure to the latest one.

Veeva also generated approximately $909 million of net income on $3.20 billion of revenue in its latest full year. This is already a large, profitable software business serving 1,552 customers.

Its R&D and Quality products are especially relevant to our question. Veeva reported roughly $1.43 billion of subscription revenue from R&D and Quality Solutions in the latest year, compared with about $906 million two years earlier. That is close to 58% growth in two years.

Drug development produces long-lived records around trials, quality, submissions, safety and regulated processes. AI can summarize those documents or help employees navigate them, but pharmaceutical companies still need controlled versions, permissions and audit trails.

Veeva itself is now explicitly pushing AI into those systems. That is worth watching because it tests the same thesis we saw in accounting and vertical operations: AI becomes more valuable when attached to the authoritative dataset.

Life sciences therefore looks exceptionally durable once a vendor has earned trust.

The catch for a new founder is obvious. Customers dislike replacing these systems for the same reasons that make them attractive businesses. Winning the first large regulated customer can be brutally difficult.

Will companies always need cybersecurity SaaS?

Companies will keep paying for cybersecurity and access control, although we would expect fewer standalone security tools as larger platforms absorb overlapping categories.

Security has a permanent underlying problem: every business continually creates users, devices, cloud accounts, applications, customer data and third-party access.

Regulators reinforce that demand. Rules such as the FTC Safeguards Rule can require access controls, monitoring, multi-factor authentication in covered circumstances, risk assessments and incident-response procedures. Public companies face separate cybersecurity disclosure and governance requirements.

But cybersecurity is one area where “always needed” does not translate cleanly into “great standalone SaaS.”

A company may permanently need identity management, endpoint protection, cloud security, access reviews and threat detection while buying several of those functions from the same vendor five years from now. AI will accelerate that consolidation because one security agent can increasingly investigate information that used to require jumping between multiple dashboards.

The safest layer is the one with authority.

A product controlling which employee can access payroll, production systems or customer data has a strong reason to exist. A small dashboard that repackages alerts from other security products has a much harder argument.

So security belongs high on the durability list, but we would rank payroll, accounting and regulated systems of record above most narrow cybersecurity SaaS ideas.

Which “boring SaaS” categories look much less safe than people think?

Standalone scheduling, invoicing, forms, simple approval apps, routine reporting, basic e-signature and lightweight CRM follow-ups look much less durable than the underlying tasks suggest.

Scheduling is the easiest trap. Businesses will schedule appointments forever, but ServiceTitan already schedules technicians, Toast manages restaurant workflows, healthcare platforms schedule patients, payroll systems manage shifts, and AI assistants can increasingly arrange meetings or jobs directly from the main operating system.

Basic invoicing has the same problem. QuickBooks creates invoices from accounting data. ServiceTitan creates them from field-service jobs. Toast derives transaction records from restaurant activity. BILL connects receivables directly to payment workflows.

Forms are even easier to commoditize. Modern AI can generate a form, database schema and workflow from a plain-English prompt in seconds.

Docusign offers a more subtle example. Electronic agreements are clearly durable: the company currently has more than 1.9 million customers. Yet Docusign itself is expanding aggressively beyond the signature into Intelligent Agreement Management.

Its latest annual filing already counted more than 25,000 IAM customers. The strategic direction makes sense because the signature alone is increasingly easy to reproduce. The deeper value comes from storing agreements, extracting obligations, controlling approvals, tracking renewals and connecting contract data to other systems.

Generic CRM work also deserves skepticism. Companies will always need customer records, but “write follow-up email,” “summarize call” and “remind salesperson to respond” are exactly the sort of activities AI can absorb into a broader CRM.

Across these categories, the recurring task can fool us into overestimating the durability of the standalone product.

The better question is whether the software owns unique state that another system would struggle to recreate. If it does not, bundling becomes a serious threat.

Does AI actually hurt the strongest boring SaaS companies?

So far, AI looks more dangerous to narrow SaaS features than to the strongest boring SaaS platforms, and the latest operating numbers reinforce that view.

QuickBooks Online Accounting grew 23% in its latest fiscal year while Intuit pushed deeper into AI accounting.

ServiceTitan kept gross dollar retention above 95% while adding AI across its operating platform. Its platform revenue grew 25% in its latest fiscal year, and net dollar retention remained above 110%.

Procore is adding AI to construction workflows while maintaining 95% gross retention and increasing its $100,000-plus ARR customers by 14% in the latest reported quarter.

Toast is explicitly combining agentic AI with payments, software and restaurant operations while locations grew 22% and ARR increased 25%.

Samsara now describes its direction as moving from visibility toward systems that can act. ARR still grew 30% in its latest fiscal year.

Veeva calls the shift toward agents a major opportunity because its AI can operate inside core systems of record and proprietary life-sciences datasets. Subscription revenue grew 17% in its latest full year.

None of this proves these companies are permanently protected from AI. The current evidence covers only the first few years of widespread generative-AI adoption.

But it is enough to reject the simplistic idea that AI immediately destroys software whose workflows can be automated.

The platforms performing best have something an external AI model needs before it can safely act: identity, permissions, business rules, historical data, integrations and the current state of the company.

Users may spend far less time clicking through those products in the future. The database underneath the agent can become even more important.

What should a founder actually build if they want durable boring SaaS?

A founder looking for durable boring SaaS should go after a recurring business obligation and own enough of the underlying record that customers would lose something important by deleting the product.

The strongest opportunities usually combine three layers.

First comes the record: employees, transactions, assets, contracts, patients, jobs, compliance controls or another dataset the business cannot casually throw away.

Then comes workflow. The software should help decide what happens next: approve the invoice, dispatch the technician, reconcile the account, renew the certificate, investigate the incident or update the employee record.

Execution makes the combination much stronger. Payroll software moves money and files taxes. BILL sends payments. Toast processes transactions. ServiceTitan turns jobs into invoices and payments. Security systems grant or revoke access.

Once software knows what is true, knows what needs to happen and can make it happen, AI becomes useful leverage rather than an existential threat.

The weakest boring SaaS ideas usually start farther up the stack. They make a form prettier, generate a report, send a reminder or save someone a few clicks. Those can still become profitable businesses, but “people will always perform this task” is a poor reason to assume the software itself will last.

For a small founder, there is another trade-off. The most durable markets often have the ugliest implementation work. Industry-specific integrations, permissions, data migration, compliance requirements and strange edge cases slow development.

Those annoyances are exactly why generic AI products have trouble replacing the software quickly.

So what boring SaaS will people always need?

The clearest answer today is payroll, accounting, AP and AR control, regulated compliance records and vertical operating systems tied to real businesses. These underlying software categories have the strongest chance of surviving even as AI rewrites their interfaces.

Payroll takes the top spot for us. ADP still has 92.1% Employer Services revenue retention and relationships that its retention data imply last about 13 years. Paycom is at 91%. Paychex still serves about 800,000 payroll customers despite much higher churn among small businesses.

Accounting is almost as strong. QuickBooks Online Accounting has gone from roughly $3.38 billion to $5.05 billion of annual revenue in two years while AI has become dramatically better at bookkeeping. That is hard evidence that automating accounting work does not automatically eliminate the accounting platform.

AP, AR and payment controls follow closely. BILL processed $371.3 billion through its platform over its latest fiscal year. Money movement keeps generating records, permissions, approvals and reconciliation work.

Regulated systems may be even stickier once installed. Veeva now produces $2.68 billion of annual subscription revenue from software embedded across life sciences, with R&D and Quality subscription revenue growing roughly 58% in two years.

Vertical operating systems round out the group we like most. ServiceTitan has kept gross dollar retention above 95%. Procore is at 95%. Toast handles $215 billion of trailing payment volume across roughly 180,000 locations. Samsara has reached $1.9 billion of ARR from physical-operations software.

Security and contract management remain durable needs too, although consolidation makes us less confident about individual subcategories.

We would be much more cautious with standalone scheduling, basic invoicing, simple forms, generic reporting, routine approvals, basic e-signature and lightweight CRM automation. People will keep doing all of those things. Increasingly, they will do them inside something larger.

The boring SaaS with the best odds of lasting is therefore software that owns an unavoidable piece of business reality: who got paid, where the money went, what the company owes, what the regulator needs, who has access, what happened to an asset, which agreement governs the relationship or what happened during the job.

AI can change almost everything about how we interact with those records. Companies still need somebody to keep them straight.

Rank	Boring SaaS category	Our durability judgment	Biggest long-term risk
1	Payroll and employment compliance	Extremely high	Vendors change; need survives
2	Accounting and financial ledger	Extremely high	Interface heavily automated
3	AP, AR and payment control	Extremely high	Bundling by banks and accounting platforms
4	Regulated compliance and records	Extremely high	Large incumbents dominate
5	Vertical operating systems	Very high	Market-specific incumbent strength
6	Fleet, asset and maintenance systems	Very high	Platform consolidation
7	Identity and access management	Very high	Security-suite consolidation
8	Contract and agreement management	High	Basic e-signature commoditizes
9	Standalone scheduling	Low as a standalone category	Absorbed into larger platforms
10	Basic invoicing, forms and reporting	Low as standalone SaaS	AI and bundling commoditize features