The problem with selling software to ‘everyone’

What makes horizontal software so seductive is that it feels infinite.

“Every company is a customer.”

And technically, that is true. A workflow platform can sell to finance, HR, procurement, legal, and operations. A collaboration product can live inside startups, banks, hospitals, governments, and manufacturers. A no-code platform can power a 10-person agency and a 200,000-person enterprise. Some investors love this story because it hints at the holy grail of venture capital: an enormous TAM.

The problem is that markets do not buy software.

People do.

A CFO does not wake up wanting “workflow automation.”

They wake up wanting fewer invoice errors.

Faster close cycles.

Less tax exposure.

Fewer people manually reconciling spreadsheets at midnight before quarter-end.

This is the paradox at the center of some of the most successful software companies of the last two decades:

Nearly every iconic horizontal software company started with an extremely narrow GTM motion.

ServiceNow did not begin by trying to digitize every workflow inside the enterprise. It started with IT service management.

Salesforce did not begin by trying to be the operating system of the customer relationship. It started with sales CRM.

Atlassian did not begin by trying to power every team’s collaboration. It started with developers.

HubSpot did not begin by trying to be the marketing-to-revenue platform for SMBs. It started with inbound marketers.

Even Amazon, long before becoming “the everything store,” started with books.

All of these companies eventually expanded horizontally — by winning one workflow, one buyer, and one adjacency at a time.

Nothing kills SaaS companies faster than excess complexity.

At Pipefy, this complexity revealed itself slowly. The company launched in 2015 with an elegant thesis: give non-technical teams a no-code way to build workflows without waiting months for IT. Purchase approvals. Vendor onboarding. Financial operations. HR processes. The product spread quickly because almost every department inside a company had some broken workflow trapped inside spreadsheets, email chains, or IT backlogs.

The template gallery went viral. Enterprises self-served onto the platform. Inbound drove growth. But inbound was also hiding a structural GTM problem. Customers were arriving already knowing the problem they wanted solved. Pipefy simply had to demonstrate how the platform could solve it.

The moment the company tried to build a real outbound motion, the cracks started to appear.

Everything suddenly became harder.

Who was actually desperate enough to buy? Which workflow created enough operational pain to justify a budget? Which buyer had both urgency and authority? Which use case was painful enough for a customer to rip out their existing workflow and change behavior?

Felipe Carvalho, who helped scale Pipefy before co-founding Camu, described the realization through a deceptively simple question posed by former Airtable CRO Seth Shaw:

“Who would be crazy not to use Pipefy?”

The shallow answer was “everyone.” The honest answer was “we don’t know.”

Pipefy spent the next eighteen months eliminating verticals it was bad at, naming the persona inside each remaining vertical, and rebuilding the sales motion around a small number of high-conviction wedges.

Pipefy’s clearest re-positioning was as the long-tail complement to ServiceNow inside IT. ServiceNow owns the catalog of large, repeatable services. Pipefy could own everything ServiceNow’s team can’t economically justify building — the long tail of internal request workflows. A salesperson could pitch this in a cold email, with a named buyer (Head of IT Operations) and an obvious adjacency (procurement, then HR, once IT was anchored).

Camu, in many ways, became the inverse reaction to everything Pipefy learned the hard way.

Before writing production code, the team ran roughly 100 customer interviews across Brazilian finance teams to identify where operational pain was highest and dissatisfaction with existing software was strongest.

One finding stood out immediately:

Out of roughly 100 finance teams interviewed, only two said they were genuinely happy with their existing tooling.

The problem already had budget. The incumbents were simply not solving it well enough.

From the broad world of finance automation, Camu picked one very specific wedge: invoice intake workflows for Brazilian companies running SAP Business One.

Roughly 7,000 companies in Brazil.

Small enough to dominate.

Large enough to get to the first $1–3M ARR.

Specific enough that outbound messaging could become extremely precise.

And adjacent to multiple workflows the company could expand into over time: payments, reconciliation, expense management, audit.

On paper, the market sounded microscopic. Finance teams were still manually processing invoices into ERPs. Matching purchase orders manually. Validating tax fields manually. Reconciling discrepancies manually. Entire operational workflows inside large companies still depended on repetitive human labor.

The market already had software. It just was not solving the problem deeply enough.

So Camu narrowed further.

Not “finance automation.”

Not even “finance automation for Brazil.”

But SAP Business One customers with painful invoice-intake workflows.

Conversion rates went from roughly 1.5% to 17% to 33%.

Because the company finally knew exactly who it was building for. That level of specificity changed everything: messaging, outbound, implementation, partnerships, onboarding, customer success, product roadmap, even engineering prioritization.

The company stopped trying to be useful to many.

It became essential to a few.

And once the ICP became brutally specific, the sales motion itself became much easier to operationalize. Three rituals started powering the system:

Once the company knew exactly who it was selling to, the team could build repeatable systems around qualification, pain articulation, and deal review.

Narrow GTM did not just improve conversion. It reduced operational chaos.

The wedge-then-platform pattern is the canonical shape of every successful horizontal-platform business of the last twenty years.

Still counterintuitive.

Startups often believe growth comes from doing more.

More segments. More geographies. More features. More use cases. More channels.

But early velocity usually comes from the opposite. The companies that scale fastest are often the companies that temporarily make themselves smaller.

Smaller ICP. Smaller promise. Smaller wedge. Smaller surface area.

Until eventually they become the monopoly inside that niche. And only then do they expand.

Three things are worth noticing about how the winners actually played this.

First, every winner started in a small, named cell of the matrix. ServiceNow with IT service management. Salesforce with the AE’s individual workflow. Slack with engineering teams inside specific orgs. Not “the enterprise.” Not “every salesperson.” A named persona inside a named team inside a named function.

Second, every winner expanded along an obvious adjacency, not into the white space. ServiceNow went from ITSM into ITOM, then HRSD, then customer service. Salesforce went from SFA into Service, then Marketing, then platform. The expansion path is rarely creative. It is the next workflow the same buyer already needs.

Third, the wedge is not a constraint. It is a compounding asset. What Tidemark Capital calls the “ServiceNow archetype” — cohort expansion running to 20x+ on the earliest customer cohort — shows up across multiple winners. The first customers who came in through the narrow door buy 20x more product over time than they bought on day one. Narrow at entry. Compounding inside.

These lessons apply to AI even more than they did to SaaS. Models are more capable than ever. Building software has never been easier. The product can technically do more than ever — which makes the temptation to position horizontally even harder to resist.

Agent for everything. AI for any workflow. Platform for any vertical.

But buyers do not buy capability.

They buy relief.

—  —  —

1. On the rise of recurring impact

For two decades, enterprise software ran on one assumption — buyers paid upfront, signed multi-year contracts, and absorbed the risk that the software might not deliver value. The agent economy is inverting that. When the product is an outcome rather than a tool — claims appealed, lawsuits defended, invoices reconciled — the vendor is no longer selling access. The vendor is selling results. And the customer has very little reason to commit upfront, because the cost of switching to a different agent next quarter is roughly zero. Recurring revenue is turning into recurring impact. The companies that win the next decade will be the ones built around delivering measurable outcomes from day one.

2. On saying no as a strategic weapon

In Camu’s first year, the company turned down a marquee RFP from a brand-name Brazilian enterprise that would have generated significant ARR but consumed engineering for years on bespoke requirements outside the wedge. Every engineer-week spent on a non-wedge customer is an engineer-week not spent making the wedge harder for competitors to dislodge. The company said no. Felipe says it was the best decision they made in the early days. Most founders cannot do this because the short-term revenue gravity is too strong. The ones who can compound for years before anyone notices.

3. On finding fast-moving water

Fast-moving water is Felipe’s version of the Pareto rule — the rare bet where a small amount of energy produces a 10x return. At Camu, fast-moving water turned out to be partnering with the right influencer with a base of qualified followers who already cared about the exact workflow Camu was selling into. Or the network of systems integrators who sell and implement SAP Business One across Brazil. The incentives were already aligned. The customer relationships already existed.

4. On the senior-hire trap

Many founders, at the scale-up stage, hire a senior enterprise salesperson from a name-brand corporate to “scale GTM.” It often fails. Felipe lived through this multiple times at Pipefy and articulates why better than anyone — building a plane is very different than flying a plane. The senior hires are great executors with a playbook. Early-stage GTM doesn’t have a playbook yet. The founder still needs to figure out why people are actually buying before anyone else can scale that motion. The smartest move at Camu was the opposite of conventional wisdom — Felipe ran sales himself, AI-augmented, until the playbook was clear. Only then would he add people.

5. On AI as a one-person sales force

Felipe is personally managing 68 sales opportunities at the same time — work that used to require an entire revenue team. The mechanic is simple. Salespeople historically spend 25% of their time selling and 75% on back-office work: CRM updates, ROI models, proposal decks, deal scoring, transcript-to-summary pipelines, objection logs. AI does the 75%. From a single call transcript, Claude can populate the CRM in full detail, generate a deal-power score, build an ROI calculator, and draft a customized proposal deck in the buyer’s own words — in five minutes. The same task at Camu used to take three people an entire afternoon. Felipe estimates 50 to 70% of what used to be back-office work is now eliminated. The result isn’t just speed — it’s that one founder, AI-augmented, now has the operational reach of a 10-person sales org.

Check out my favourite moments in this reel that went viral on Instagram.

—  —  —

1. Good Strategy Bad Strategy by Richard P. Rumelt

One of the best strategy books I’ve ever read.

Most companies think strategy is: ambition, goals, vision statements, buzzwords, “we want to be the leader in…”

Rumelt argues real strategy is much less glamorous: diagnosing the actual problem, identifying the critical bottleneck, and concentrating resources against it.

The book completely changes how you think about focus, trade-offs, positioning, and organizational clarity.

Especially relevant in the AI era where everyone wants to do everything at once.

2. Gavin Baker on Invest Like the Best — Watts and Wafers

The central frame — watts and wafers — names the two physical constraints that will dictate the next phase of AI: power supply and TSMC capacity. Baker thinks the near-term power shortage starts to ease in 2027–2028 as new energy sources come online, and that orbital compute solves it in the long term. On wafers, he argues TSMC’s capacity discipline is the single most important variable to watch — and the cleanest indicator of whether AI infrastructure is in a bubble. Also worth the listen for the segments on Elon’s Terrafab, the disaggregation of GPUs, and whether frontier-model economics keep accruing to the labs.

Thanks for reading,

Olga 

P.S. If this issue was valuable to you please share it with a founder who needs to hear it. Let’s build LATAM’s next tech leaders—together

🎙 The J Curve  is where LATAM's boldest founders & investors come to talk real strategy, opportunity and leadership.