Business Owners Are Walking In Saying "Can You Build This?" and We Say Yes.

Almost every Heed discovery call now starts with a sentence the business owner did not used to be allowed to say. "I wish my CRM did X. Can you build it?" Or, "Our intake process is broken. Can you build something that just handles it?" Or, the most common one, "We have this thing we do every Friday that takes the whole team three hours. Can you make it go away?"

Five years ago, the honest answer to all of those was the same. Yes, but not for that budget. Custom software for an SMB started at six figures, took quarters, and usually got delivered as a thing the team did not actually want by the time it shipped. The right answer was almost always "use a SaaS product and bend your workflow around it."

That is not the answer anymore. The answer is now yes.

What Owners Are Actually Asking For

Three real-shape requests we have built or are building in the last two quarters. None of them would have been viable for an SMB in 2021. All three are normal projects in 2026.

A field-team mobile app. An owner of a multi-state field services business walked in with a phone full of photos. He showed us the daily report process. Crew lead takes photos at the jobsite, texts them to the manager, the manager forwards to the controller, the controller types up a daily report, the owner reads it the next morning and asks the same five questions every time. He wanted: an app the crew uses on site that takes the photos, runs computer vision against the firm's history of similar jobs, identifies the categories of damage or work performed, and auto-fills the daily report. We are building it. The vision model handles the categorization. The crew lead does a 30-second review and submits. The controller does not type anymore. The owner reads cleaner reports.

An employee dashboard for a single legal practice area. A multi-attorney estate and family law firm serving high net worth families wanted a custom dashboard that triaged inbound matter intakes specific to their practice areas (living trusts, conservatorships, special needs trusts), pulled the right intake template, drafted the engagement letter from the firm's voice, and routed the matter to the right attorney based on practice area and load. We built it. It runs on Anthropic Claude for the routing decisions, OpenAI for image generation in marketing materials, and Perplexity for case-law research.

A custom landing-page builder branded to a single business. A regional service business wanted to spin up location-specific landing pages for paid ad campaigns without paying a marketing agency $300 per page. We built them a builder. AI copywriting trained on the firm's existing voice and offers. Their team picks a template, fills in three fields, and the page goes live on their domain. The cost per page collapsed.

What Changed

None of these are technically novel. Computer vision has existed for years. Routing logic has existed for decades. Landing-page builders are a saturated category. What changed is the cost and the time.

Cloudflare Workers, Hyperdrive, Pages, and R2 collapsed the infrastructure cost of standing up a custom backend. The Anthropic API and the OpenAI API made enterprise-grade AI a line item, not a project. Claude Code and MCP servers compressed the engineering loop. Cloudflare Zero Trust gave us the security layer without a separate identity vendor. AI-fluent developers, what people sometimes call vibe coders, ship in days what teams used to plan in quarters.

We have leaned into this on the team side. We are actively hiring AI-fluent builders in part because demand is clearly outpacing the supply of developers who know how to work this way. The category is real and it is growing.

What We Say Yes To

The pattern that gets a yes is consistent across industries. Three things have to be true.

First, the math has to work. We run the ROI model on every engagement before we propose a build. Hours saved times blended rate has to come out clean against the build cost in 12 months or less. Without that, the project is interesting but not worth doing.

Second, the workflow has to be stable enough to automate. If the rules of how the work gets done change every quarter because the business is still figuring itself out, custom software is not the right move yet. Document the workflow first. Live with the documentation for a quarter. Then build.

Third, the team has to have the bandwidth to maintain what gets built. We deliver the dashboard or the app, but it lives inside the client's business. Someone has to own it on their side. That can be a fractional CAIO relationship, an internal owner, or a maintenance retainer with us. Without it, the build degrades.

What We Say No To

We say no to projects that require continuous bespoke human judgment to even define the rules. If the workflow cannot be described as "if this happens, do that" with a manageable number of branches, AI is not yet a good fit. Some legal work falls in this category. Some creative work. Some senior advisory work. The right answer there is to give the human the best tools and not try to replace the judgment.

We say no when the math does not work. Two hours of weekly time saved against a $30,000 build cost is not a project. We will tell you so on the discovery call.

We say no when the workflow is genuinely upstream of a strategy decision the business has not made. If the owner is not sure whether the future of the firm is more service A or more service B, building software optimized for either one is premature. We will recommend a strategy conversation first, often a fractional CAIO engagement.

The POC-First Approach

Every "can you build this" conversation that survives the math turns into a Proof of Concept inside two to four weeks. Working code. Real data. The team can use it. We do not propose a $50,000 build off a slide deck. We propose a $5,000 to $15,000 POC, deliver it, and let the team decide whether to scale it.

That keeps the risk close to the ground and makes the "yes" answer cheap. If the POC works, the full build follows. If it does not, the client paid for a working prototype and learned something concrete about their workflow. Either outcome is better than a six-month plan-and-build with a binary outcome at the end.

Five years ago, "can you build this" was a question that ended with "not for that budget." Now it ends with "let us run the POC and see." If you have been holding a question like that, bring it. The answer might be yes.