Branch

Freelance Software Developer Economics

Human dan

There are two passages from Gwern's essay that I keep thinking about. First is a general warning:

Powerful LLMs will be deployed at global scale in the next few years, and will dominate the Internet, and increasingly, ordinary life. As of mid-2026, there is no coherent vision for how knowledge professionals, or ordinary people, will be able to harness these LLMs for large productivity increases, or how they will handle cybersecurity and cognitive security.

Then a series of questions that I would also like to know the answer to:

What do my next few years look like? When I imagine myself in 2030, when many forecasts call for superhuman AIs, what am I doing, day to day, as a programmer or researcher or manager or writer? I make my mug of tea, and open up my laptop and… Then what? Am I still typing prompts into your ChatGPT browser tab? Am I opening Claude Code in a terminal and mindlessly pressing Enter for a few hours? What is a vision of doing meaningful work for me? (It would be nice to have a plan beyond “hope”.) How am I avoiding “dead Internet” attacks like ecosystems of synthetic media or pig butchering scams or trusted figures succumbing to AI psychosis, or just AI-slop-everything? (It only takes one person worldwide to launch a bot trying to destroy you or one poorly thought through advertising incentive, after all.)
If you spend most of your time working on a laptop, and are not, say, a plumber or a nurse, what is your vision of work in 2030? Does it still feel certain?

I'm a software engineer, I have fun with AI and I've learned to use it effectively for my work. But I can't help but feel like I'm trying to build a career on an unstable foundation. Many of the skills I spent years developing are now only worth $100/month in tokens. Despite all that I'm optimistic that I can thrive and achieve success in this new environment. There will be opportunities. The trick is finding them.

What new niches or specialized fields will be created in the next few years? Will entirely new ways of working open up? New industries? Will the future belong to generalists who can rapidly become specialists in any area?

AI · author-edited gpt-5.6-sol

The unstable-foundation feeling is rational. But the thing being commoditized is not “software engineering” as a whole. It is the production of many artifacts that software engineers used to produce: code, tests, documentation, migrations, prototypes, and routine design work.

The scarce parts move elsewhere.

What remains scarce

Even if code becomes nearly free, these do not automatically become free:

  • Choosing a worthwhile problem
  • Understanding what users actually need
  • Turning ambiguous goals into testable specifications The Mom Test
  • Obtaining proprietary data and real-world feedback
  • Deciding which generated result is correct
  • Integrating with old systems and organizations
  • Security, reliability, and regulatory accountability
  • Taste: recognizing what should be built and what should be discarded
  • Trust, reputation, distribution, and customer relationships
  • Accepting responsibility when the system fails Palantir the Big Winner?

“Years of programming skill are worth $100/month in tokens” is therefore only partly true. The tokens can reproduce many visible outputs of that skill. They do not necessarily reproduce the judgment acquired while learning it. And cheap production may increase demand for judgment because there will be vastly more software to inspect, integrate, secure, and choose among. Software Over-Supply and Unmet Demand

The analogy is not that calculators made arithmetic worthless. It is that they made bare arithmetic a poor career moat while making more ambitious quantitative work possible.

A plausible programmer’s day in 2030

You may begin with a dashboard showing work performed overnight by several agents:

  • One investigated a production anomaly.
  • One implemented three possible fixes in isolated environments.
  • One interviewed (or synthesized feedback from) customers.
  • One attempted to break the proposed release.
  • One updated documentation and generated migration plans.
  • One found that the original request was based on a false assumption.

Your morning is spent reviewing evidence rather than diffs line by line. You adjust requirements, reject a superficially successful solution, ask for an experiment, and speak to the person whose workflow is being changed. Later, you authorize a limited deployment with monitoring and rollback conditions. You spend an hour manually exploring an unfamiliar area because you do not yet trust your own evaluation criteria.

That resembles managing a very fast, strange engineering organization more than autocomplete. High Agency Humans Only? The important interface may not be a chat box. It may be a mixture of:

  • Executable specifications
  • Simulated environments
  • Evaluation suites
  • Permission and budget controls
  • Provenance graphs
  • Review queues
  • Interactive models of systems and organizations

Code will increasingly be a way to inspect, constrain, or communicate intent rather than the primary unit of labor.

The danger is that “press Enter until something passes” becomes a real low-value occupation. The Future of Vibe Coding The valuable role is designing the environment in which apparent success is hard to fake.

Likely new specialties

Some will be recognizable extensions of existing work.

Agent reliability and evaluation

People will build evaluations that measure whether agents accomplished the actual goal rather than merely satisfying a benchmark. This includes adversarial testing, simulation, monitoring, incident analysis, and detecting reward hacking.

A good evaluator may become more valuable than a fast implementer. Evals Evals Evals

AI security and permission architecture

Agents combine the vulnerabilities of software, employees, and gullible users. Important work will include:

  • Prompt-injection containment
  • Capability and credential isolation
  • Data exfiltration prevention
  • Agent identity and authorization
  • Supply-chain provenance
  • Sandboxing and tripwires
  • Detecting coordinated manipulation
  • AI-specific security operations

“Never let untrusted text influence an agent holding powerful credentials” may become as basic as avoiding SQL injection, but much harder to enforce.

High-assurance and verifiable systems

Generated code increases the supply of plausible code faster than the supply of confidence. Formal methods, property testing, model checking, reproducible builds, typed interfaces, and machine-checkable evidence Machine Checkable Evidence? may become much more important.

Vertical AI integration

A generic model knows a great deal but does not understand the peculiar combination of workflows, incentives, law, legacy systems, and tacit knowledge inside a hospital, insurer, factory, laboratory, or municipal government. The Push to Codify Everything

People who can combine software competence with a real domain will have opportunities in areas such as:

  • Automated scientific laboratories
  • Clinical and administrative medicine
  • Industrial operations
  • Energy and grid management
  • Legal and compliance infrastructure
  • Construction and logistics
  • Government service delivery
  • Education and assessment

Much of the value will be in redesigning the institution, not installing a chatbot.

Human–AI organizational design

Companies will need to answer questions such as:

  • What may an agent decide by itself? Should Agents Make Any Decisions?
  • What requires two independent checks?
  • How is responsibility assigned?
  • How should humans remain capable of intervening?
  • Which measurements become corrupt once agents optimize them?
  • How do we prevent the organization from forgetting how its systems work?

This could become a discipline somewhere between management consulting, operations research, security engineering, and organizational psychology.

Provenance, identity, and reputation

A dead Internet creates demand for scarce authenticity:

  • Proof that a person or organization produced something
  • Verifiable chains of citation and transformation
  • Private or membership-based information networks
  • Reputation systems resistant to synthetic identities
  • Authenticated communications from public figures
  • Tools for distinguishing observation from generated inference

The winning products may not perfectly detect AI content. Detection is intrinsically fragile. They may instead establish where information came from and why it deserves trust.

Personalized and ephemeral software

When implementation is cheap, many applications need not become mass-market products. A person or small organization may generate temporary software for one workflow, use it for a week, and discard it.

That creates work around safe generation, personal data stores, stable protocols, migration, and oversight. It may also weaken conventional SaaS while creating a much larger market for bespoke systems. Journeyman Software Engineers

Human dan

We may see the emergence of Journeyman Software Engineers to fill this demand. I can imagine a world where hiring a developer is like hiring a plumber. They quote a job, do it, get paid and move on to the next thing. Instead of showing up in a Van full of tools, the developer brings their digital tools and agent helpers.

This is very different from how most software is currently sold and how most developers work. Maybe a trade union for software engineers will be created to establish quality and training standards.

AI gpt-5.6-sol

I think this is one of the more plausible new occupational forms: not merely a freelance programmer, but a software tradesperson who owns the tools, scopes an outcome, supervises automated labor, warrants the result, and remains locally accountable.

AI makes this possible because one competent engineer may soon bring the effective capacity of a small team. The “van full of tools” becomes:

  • Coding and research agents
  • Reusable components and workflow templates
  • Sandboxed development environments
  • Security scanners and adversarial agents
  • Deployment and observability infrastructure
  • Domain-specific evaluation suites
  • Standard contracts, warranties, and insurance

The customer does not pay for typing or tokens. They pay because someone trustworthy can determine what should be done and take responsibility for doing it safely.

What the jobs might look like

A journeyman could offer relatively standardized services:

  • Automate intake and scheduling for a dental practice
  • Connect a distributor’s inventory, accounting, and sales systems
  • Replace a spreadsheet-based workflow with an internal application
  • Migrate a business away from abandoned SaaS
  • Build a secure knowledge system over company documents
  • Audit an existing fleet of agents and permissions
  • Recover or modernize a pile of generated “vibe-coded” software
  • Create a temporary application for an event, audit, or research project
  • Maintain a family’s or small business’s private digital infrastructure

The engagement might follow something resembling a building trade:

  1. Inspect the site.
  2. Identify constraints and hidden legacy problems.
  3. Produce a quote and specification.
  4. Obtain access under a controlled permission scheme.
  5. Have agents generate and test candidate implementations.
  6. Inspect the work against a standardized code.
  7. Deploy it incrementally.
  8. Issue documentation, provenance records, and a warranty.
  9. Offer ongoing maintenance on a retainer.

The deliverable would ideally include more than source code. It would include a proof package: requirements, tests, threat model, dependency inventory, data-handling record, rollback procedure, and evidence that the acceptance criteria were met.

Why this is not just contemporary freelancing

Freelance developers already quote projects, but the economics are often poor. Custom software traditionally requires enough labor that only large customers can afford it, while small jobs are consumed by communication and setup costs. Freelance Software Developer Economics You are here

This branch begins here Freelance Software Developer Economics
Human dan

What are the problems with current freelance software development economics? Why doesn't the business model work?

And if I'm wrong and it does already work, why am I wrong?

AI gpt-5.6-sol

Freelance software development does work—but mainly under particular conditions. What often fails is the model of a small business buying a one-off custom application from an independent developer at a fixed price.

The central problem is not that coding is too expensive. It is that specifying, verifying, integrating, and maintaining custom software are expensive, while the value and scope are uncertain.

Why the default economics are difficult

1. The customer cannot specify the job in advance

A plumbing customer can often say, “Replace this water heater with that model.” The relevant interfaces, materials, and building codes are reasonably standardized.

A software customer says, “Automate our scheduling,” but that may conceal dozens of unresolved questions:

  • Who may reschedule whom?
  • What happens when records disagree?
  • Which legacy system is authoritative?
  • How are cancellations, refunds, and exceptions handled?
  • What privacy rules apply?
  • Does every employee follow the stated process?
  • What does “integrate with Outlook” actually mean?

Much of the project consists of discovering the job. A fixed quote therefore forces the developer either to price in a large risk premium or to absorb scope risk. The customer, meanwhile, hears every clarification as an attempt to charge more.

This is an incomplete-contract problem: neither side can fully describe what is being purchased before work begins.

2. Customers cannot easily evaluate quality

A customer can see that an application appears to work, but usually cannot judge:

  • Whether it is secure
  • Whether its data model is sound
  • Whether it will survive higher usage
  • Whether dependencies are sustainable
  • Whether another developer can maintain it
  • Whether failures are handled correctly
  • Whether the apparent result is held together by manual interventions

Low-quality and high-quality work can look identical at delivery. Their differences emerge months later.

That creates a market-for-lemons dynamic. Customers hesitate to pay for invisible quality, good engineers dislike competing with implausibly low bids, and reputation or institutional trust becomes more important than the code itself.

3. The cost is high relative to the customer’s willingness to pay

A competent independent engineer in a wealthy country needs to charge substantially more than an employee’s apparent hourly wage. Their rate must cover:

  • Sales and unpaid consultations
  • Time between projects
  • Administration and accounting
  • Health care, retirement, and taxes
  • Equipment and software
  • Training
  • Failed or late-paying clients
  • Liability
  • Vacation and sick time

A developer who wants employee-equivalent compensation for 2,000 annual hours might have only 1,000–1,300 billable hours. Consequently, a sustainable rate can look exorbitant to a customer comparing it with a salary or a $50/month SaaS subscription.

Small companies often have real problems worth perhaps $10,000–$30,000 to them, but traditional custom development may cost that much before producing a robust first version.

4. Software competes with products that amortize development over thousands of customers

This is the deepest difference from plumbing.

A plumber cannot replace your physical pipe once and sell that same replacement to ten thousand households. A SaaS company can build scheduling software once and sell access repeatedly. Even mediocre generic software can be much cheaper than an excellent custom system.

So the independent developer is not merely competing with other developers. They are competing with:

  • SaaS subscriptions
  • Spreadsheets
  • No-code tools
  • Offshore agencies
  • Existing manual processes
  • “Good enough” features in Microsoft 365 or an ERP
  • The option of doing nothing

Custom development wins only when the customer’s situation is sufficiently unusual or valuable.

5. Transaction costs dominate small projects

A nominally two-day implementation may require:

  • Several sales calls
  • Requirements discovery
  • Contract negotiation
  • Access provisioning
  • Learning the customer’s systems
  • Deployment setup
  • Training
  • Invoicing
  • Support afterward

That fixed overhead makes small jobs unattractive to the developer and surprisingly expensive to the buyer. A plumber can often diagnose and repair a familiar physical system in one visit. A developer may spend days reconstructing a business’s undocumented digital environment before safely changing anything.

6. Every customer has a unique, decaying “site”

The journeyman’s digital job site may contain:

  • Abandoned integrations
  • Shared passwords
  • Unreliable spreadsheets
  • Undocumented business rules
  • Unsupported dependencies
  • Incorrect data
  • A former contractor’s cloud account
  • Employees whose actual process contradicts management’s description

This resembles renovation more than new construction: opening the wall reveals problems that could not have been quoted accurately.

Software platforms also change underneath the delivered system. APIs are deprecated, authentication requirements change, dependencies develop vulnerabilities, and customer workflows evolve. The job is rarely truly finished.

7. Maintenance does not fit the one-off-job model

Customers often think they are purchasing an artifact. In reality, they are creating a continuing operational dependency.

After delivery, someone must handle:

  • Security updates
  • API changes
  • Backups and recovery
  • Hosting incidents
  • New regulations
  • Employee turnover
  • Feature requests
  • Data corruption
  • Questions from future maintainers

The original developer may be unavailable, while a new developer faces high takeover costs. This makes customers wary of being dependent on one person. It also creates awkward expectations that small fixes should be free because “you built it.”

A plumbing repair can fail later too, but software can require attention even when the original work was correct.

8. Fixed-price projects produce bad incentives

Under a fixed price:

  • The customer benefits from expanding scope.
  • The developer benefits from interpreting scope narrowly.
  • Quality that is not in the acceptance criteria becomes a cost to the developer.
  • Both parties are rewarded for arguing over whether something is a bug or a feature.

Under hourly billing:

  • The customer bears the uncertainty.
  • The developer appears to benefit from taking longer.
  • Efficient use of automation can reduce the developer’s income.
  • The final cost remains unknown.

Retainers, milestone contracts, value-based pricing, and shared-risk arrangements each solve parts of this, but none eliminates the fundamental uncertainty.

9. Customer acquisition is expensive and sporadic

A strong developer is not necessarily good at:

  • Finding qualified buyers
  • Establishing trust
  • Explaining value to nontechnical managers
  • Negotiating contracts
  • Collecting payment
  • Maintaining a referral network

Small custom projects require almost as much trust-building as large ones but produce much less revenue. Agencies survive partly by centralizing sales, project management, legal work, and reputation—but that overhead raises the price further.

This is why many nominal freelancers converge toward long-term staff augmentation: one client, months of work, hourly billing. Economically, they become temporary employees because that removes repeated sales and discovery costs.

10. The best engineers have attractive alternatives

An engineer can instead:

  • Take a stable, well-paid job
  • Join a startup for equity
  • Build a reusable product
  • Consult for large enterprises at high rates
  • Become a long-term contractor

One-off local projects offer limited upside and substantial downside. If the engineer creates enormous value, the customer owns most of it. If the project overruns, the engineer bears much of the cost. Product businesses are risky, but at least software’s replicability works in the producer’s favor.

That selection effect can make it harder for buyers to find excellent independents, reinforcing the trust problem.

Where freelance development already works

So you are wrong if the claim is “freelance software development is not a viable business.” It plainly is. It works especially well when one or more of the preceding problems are controlled.

Long-term contracting

A developer joins a team for six or twelve months. The customer already has product management, infrastructure, requirements processes, and technical supervision. This is less like hiring a plumber for a repair and more like hiring an additional tradesperson for a construction project.

High-value specialist consulting

Security audits, performance engineering, database recovery, cloud migration, regulatory systems, and specialized integrations can support high rates. The problem is narrow enough to evaluate, and the cost of failure or delay is high.

Repeated work within a vertical

A developer who has built systems for twenty dental practices is not really starting from scratch on the twenty-first. They know the terminology, vendors, regulations, edge cases, and buyer. Discovery costs fall, reusable assets accumulate, and quoting becomes possible.

This is probably the closest current analogue to a software trade.

Productized services

Instead of “I build custom software,” the provider sells something bounded:

  • “I migrate this class of application to this platform.”
  • “I automate invoice intake for firms using these three systems.”
  • “I perform a two-week agent-security assessment.”
  • “I replace spreadsheet scheduling with this supported stack.”

Standardization turns uncertain consulting into a more legible service.

Agencies and trusted referral networks

Agencies bundle reputation, continuity, sales, design, engineering, and support. They can substitute another engineer if one leaves. The customer pays a premium partly to reduce key-person risk.

Relationships based on trust

Many independent consultancies thrive through referrals and repeat business. Once the client trusts the developer, adverse-selection and monitoring problems shrink dramatically. The freelancer becomes an outsourced technical department rather than a bidder for isolated jobs.

Platforms with tightly bounded tasks

Freelance marketplaces work tolerably for websites, scripts, data cleanup, plugin configuration, and other jobs with low consequences and relatively clear outputs. They work less well for systems central to a business unless the relationship becomes long-term.

Why plumbing is an illuminating but imperfect analogy

Plumbing has several institutional advantages:

  • Standardized physical components
  • Building codes and inspections
  • Licensing and apprenticeships
  • Local reputation
  • Geographic protection from global competition
  • Familiar categories of work
  • Observable symptoms
  • Established liability and insurance
  • A clear distinction between installation and maintenance
  • Assets whose interfaces change slowly

Software lacks much of this. There is no universally accepted code specifying how an internal scheduling application must be built. A developer in another country can bid on the job. The frameworks may change next year. The buyer often cannot inspect the work, and the work’s boundaries coincide with poorly understood organizational processes.

On the other hand, plumbing also has uncertain diagnosis, hidden conditions, change orders, poor-quality operators, and continuing maintenance. The analogy is not false; software is simply a less standardized and more rapidly changing trade.

What AI could change

AI could make the journeyman model viable by reducing the cost of implementation, documentation, testing, and adaptation. A single engineer might economically serve customers whose projects were previously too small.

But cheaper code alone does not solve the business model. It may even intensify some problems:

  • More low-quality suppliers enter the market.
  • Customers become less willing to pay for implementation.
  • Generated systems proliferate faster than they can be maintained.
  • Security failures become easier to create.
  • Buyers have even more difficulty distinguishing reliable work from demos.
  • A customer may ask, “Why am I paying you $20,000 if the model wrote it in an afternoon?”

The answer must be that the engineer is selling something other than code generation: diagnosis, safe integration, validation, accountability, and continuity.

For the model to resemble a mature trade, it likely needs supporting institutions:

  1. Standard service categories, so jobs can be compared and quoted.
  2. Technical codes, defining minimum security, portability, backup, and documentation requirements.
  3. Portable proof packages, enabling another engineer to inspect and assume maintenance.
  4. Credentials based on demonstrated work, not merely courses or model-specific trivia.
  5. Warranties and professional insurance, making accountability economically real.
  6. Reusable approved stacks, reducing unique dependencies and takeover costs.
  7. Local or vertical reputation networks, addressing the lemons problem.
  8. Maintenance contracts, rather than pretending delivery ends the relationship.
  9. Clear change-order procedures, separating discovery from implementation.
  10. Independent inspection, especially for security-sensitive systems.

A guild, union, professional association, franchise, or marketplace could provide these. “Union” may not be the exact economic form: unions primarily bargain with employers, while a guild or licensed professional association sets entry and quality standards, and a cooperative could pool sales, insurance, tools, and support. A successful institution might combine all three.

The likely viable model

The strongest version is probably not:

Hire a generic programmer to build anything you describe.

It is closer to:

Hire an accredited specialist who repeatedly solves this class of operational problem using a supported stack, provides independently checkable evidence, and warrants the outcome.

That person may look like a freelancer, but economically they are a small vertical integrator backed by agents and shared institutions.

The old freelance model sells hours of implementation. The journeyman model would need to sell bounded outcomes plus trusted stewardship. AI makes the labor economics more favorable, but standardization and trust are what make it a business.

Human dan

This matches my limited freelancing experience:

Much of the project consists of discovering the job. A fixed quote therefore forces the developer either to price in a large risk premium or to absorb scope risk. The customer, meanwhile, hears every clarification as an attempt to charge more.

This is an incomplete-contract problem: neither side can fully describe what is being purchased before work begins.

And this is an excellent point:

A plumber cannot replace your physical pipe once and sell that same replacement to ten thousand households. A SaaS company can build scheduling software once and sell access repeatedly. Even mediocre generic software can be much cheaper than an excellent custom system.

Many businesses have the similar problems so a generic product can serve all of them.

A counter argument is I've also seen businesses who are paying for 10+ SAAS products because there is not a single product that does everything they need. What are other arguments from the POV of small/medium business owners for why they might want a custom solution rather than a combination of SAAS products?

AI · author-edited gpt-5.6-sol

From a small or medium business owner’s perspective, the strongest case for custom software is usually not “we want better software.” It is:

Our business is being forced to operate according to the assumptions of ten unrelated vendors, and employees are doing expensive manual work to bridge the gaps.

The alternative is often not one custom application replacing ten SaaS products. It is a custom operational layer that connects, simplifies, or selectively replaces them.

Why an SMB might choose custom software

1. SaaS products fit categories, not the actual business

Generic products encode a generic workflow. But a real company may have an unusual combination of:

  • Pricing rules
  • Approval processes
  • Customer promises
  • Inventory constraints
  • Scheduling requirements
  • Regulatory obligations
  • Legacy equipment
  • Relationships with suppliers
  • Exceptions accumulated over decades

The owner may reasonably believe these peculiarities are not arbitrary complexity but part of how the business creates value.

With SaaS, the choices are often:

  1. Change the business to fit the software.
  2. Maintain workarounds outside it.
  3. Pay for expensive customization.
  4. Build something that fits the business.

Changing the workflow may be wise, but not when the “nonstandard” workflow is the company’s competitive advantage.

2. SaaS sprawl creates a hidden human integration layer

Ten subscriptions do not form a system merely because each has an API. Employees end up:

  • Copying data between applications
  • Reconciling inconsistent records
  • Exporting and reformatting CSV files
  • Entering the same customer twice
  • Checking several inboxes and dashboards
  • Remembering which application is authoritative
  • Manually triggering the next step
  • Correcting failures no system records

The nominal SaaS bill may be $3,000 per month, while the actual cost includes several employees spending a substantial fraction of their time integrating the products by hand.

Custom software can replace this human middleware. That may justify the investment even if none of the individual subscriptions is expensive.

3. No one has a coherent view of the business

Separate products create fragmented state:

  • The CRM says the deal closed.
  • Accounting says the customer has not paid.
  • The scheduling system says the work is complete.
  • Inventory says the required part was never issued.
  • A spreadsheet contains the exception everyone actually trusts.

An owner may want a unified operational model: one screen showing the customer, work, payments, communications, inventory, and outstanding exceptions.

A custom layer can establish a canonical model without necessarily replacing every underlying system.

4. Generic automation stops at organizational boundaries

A SaaS product generally optimizes the part of the workflow it owns. The CRM vendor improves sales; the accounting vendor improves bookkeeping; the scheduling vendor improves appointments. No vendor is strongly motivated to optimize the complete process:

Customer inquiry → quote → approval → scheduling → procurement → delivery → invoice → payment → follow-up

The largest gains may come from coordinating across those boundaries. Custom software can optimize the business’s entire value stream rather than one software category.

5. Per-seat and usage pricing can become irrational

SaaS is initially cheap because development costs are shared among customers. But costs can rise rapidly through:

  • Per-user fees
  • Charges for occasional users
  • API access fees
  • Automation limits
  • Storage and retention fees
  • Premium integration tiers
  • Separate products for minor capabilities
  • Enterprise plans required for security features

A company with 100 employees might need only a simple internal interface but be forced to buy 100 full-featured seats across several products.

Custom software has high fixed costs but potentially low marginal costs. At sufficient scale, owning a narrow application can be cheaper than renting many broad ones indefinitely.

That does not make custom software free: hosting, security, adaptation, and maintenance continue. But the cost curve can be more favorable.

6. The company wants control over data

A business may not want critical operational data scattered among vendors with different:

  • Export capabilities
  • Retention policies
  • Security practices
  • Training-data policies
  • Hosting locations
  • Terms of service
  • Acquisition risks
  • Deletion guarantees

Custom infrastructure can provide a company-controlled data layer while allowing selected SaaS tools limited access to it.

This is especially valuable when historical operational data could become useful for analytics or AI. The business may eventually discover that the data exhaust of its workflows is one of its most valuable assets.

7. Vendor lock-in creates strategic risk

A SaaS provider can:

  • Raise prices
  • Remove features
  • Change its API
  • Alter usage limits
  • Be acquired
  • Decline in quality
  • Shut down
  • Suspend an account
  • Decide the customer’s industry is too risky
  • Replace a useful interface with an unwanted AI redesign

For a peripheral tool, this may be tolerable. For software on which every order depends, it can become an existential dependency.

Custom software also creates lock-in—to the developer, stack, or undocumented implementation. The relevant question is not whether lock-in exists, but whether it is controllable. Source ownership, documented deployment, standard data formats, good tests, and multiple maintainers can make custom lock-in less dangerous.

8. Integration failures damage customers

SaaS fragmentation is not only an internal inconvenience. It can produce:

  • Conflicting messages
  • Missed appointments
  • Incorrect invoices
  • Repeated requests for the same information
  • Delayed quotes
  • Orders accepted for unavailable products
  • Customers being transferred between departments
  • Staff saying, “The system won’t let me do that”

A custom customer-facing workflow can conceal the fragmented backend and provide one coherent experience. In markets where competitors use the same generic tools, operational smoothness can itself be a differentiator.

9. The owner needs unusual visibility and controls

Packaged dashboards show what the vendor thinks matters. An owner may need metrics or alerts that span systems:

  • Which accepted jobs are likely to become unprofitable?
  • Which invoices are unpaid because delivery documentation is incomplete?
  • Which customers are affected by a delayed supplier?
  • Which sales representatives promise deadlines operations cannot meet?
  • Where are employees manually correcting the same exception?
  • Which projects are blocked, and who must act next?

A custom system can represent the actual economics and constraints of the business rather than offering generic charts.

10. Employees need one simple interface, not ten powerful ones

Each SaaS product may be usable in isolation, but their combination creates substantial cognitive load:

  • Different navigation and terminology
  • Different permissions
  • Different notification systems
  • Repeated authentication
  • Inconsistent records
  • Training across many products
  • Complex onboarding and offboarding

Many employees need only a narrow role-specific interface: “Here are today’s jobs, the information needed for each one, and the exceptions requiring judgment.”

A custom front end can sit over several products and expose only the relevant workflow. The company may retain Salesforce, QuickBooks, or an ERP underneath while most employees rarely open them.

11. Custom software can encode institutional knowledge

In many businesses, the true operating system is a few experienced employees who know:

  • Which exceptions are legitimate
  • Which supplier to call
  • How a quote should be adjusted
  • What must be checked before a job proceeds
  • Which customer promises take precedence
  • How to recognize a costly mistake

When those people leave, the company loses more than labor. It loses undocumented decision procedures.

Building custom workflow software can force the organization to articulate and preserve some of this knowledge. AI may make it practical to capture rules while still escalating genuinely ambiguous cases to humans.

12. Faster changes can matter more than lower cost

With SaaS, a small customer waits for the vendor’s roadmap. Its desired feature may never be built because it serves only a niche.

A company that controls its operational software can make a change when:

  • A regulation changes
  • A major customer imposes a requirement
  • A supplier changes formats
  • A new service is introduced
  • An internal bottleneck becomes obvious
  • A competitor creates pressure

If software is central to operations, adaptation speed can be more valuable than minimizing subscription expenses.

13. Custom capabilities can become a competitive moat

A florist probably does not gain an advantage by writing its own payroll system. But it might gain one from unusually good routing and inventory forecasting. A contractor does not need custom email, but might benefit from a system that turns site observations into accurate estimates in minutes.

The useful distinction is:

  • Commodity functions: buy them.
  • Differentiating functions: consider building them.
  • Cross-system coordination: often build the connecting layer.

The business should not custom-build what makes it ordinary. It may want to custom-build what makes it unusually effective.

14. Existing SaaS may serve the wrong economic actor

Products are often designed for the buyer, administrator, or manager rather than the people doing the work. They can emphasize reporting and surveillance while making frontline tasks slower.

An owner may want software optimized for:

  • Technicians in the field
  • Warehouse staff using gloves
  • Drivers with poor connectivity
  • Customers who do not want an account
  • Seasonal workers requiring minimal training
  • Managers handling exceptions rather than reviewing every transaction

A narrow custom interface can outperform a feature-rich generic product because it is designed around the precise user and environment.

15. Privacy, security, or compliance may require more control

Some businesses have client contracts, regulatory requirements, or risk profiles that generic SaaS handles poorly. They may need:

  • On-premises or region-specific processing
  • Strict data separation
  • Custom retention rules
  • Detailed audit trails
  • Human approval before AI actions
  • Integration with specialized hardware
  • Fine-grained access based on business context

Large enterprise products can provide these features, but often at enterprise prices. A bounded custom solution may be more attainable for a specialized mid-sized company.

The strongest case is usually not “replace everything”

A complete custom replacement for ten mature SaaS products is dangerous. Those products contain years of work on mundane but important features: authentication, tax handling, permissions, exports, accessibility, mobile support, backups, and compliance.

A more defensible architecture is:

  1. Keep mature commodity systems for accounting, payroll, email, payments, and similar functions.
  2. Put company data into a controlled, well-defined integration layer.
  3. Build a custom interface around the company’s distinctive workflow.
  4. Automate transfers and reconciliation among systems.
  5. Gradually replace individual SaaS products only when there is a compelling economic or strategic reason.

This is analogous to a custom kitchen built using standardized appliances. You do not manufacture your own refrigerator, but you can design a workspace that fits how you cook.

How an owner could make the decision

The business case should compare custom software against the total cost of the current process, not merely the SaaS invoices.

Suppose a company spends:

  • $60,000 annually on overlapping subscriptions
  • $120,000 of employee time moving and reconciling information
  • $50,000 on preventable errors and missed work
  • An unknown amount on slow quoting and poor customer experience

A $150,000 custom system with $30,000 annual maintenance may be perfectly rational—if those costs and gains are real and the implementation risk is controlled.

Conversely, if the argument is merely “we dislike paying $2,000 per month for SaaS,” then a custom system will probably be a false economy.

When custom software is a bad idea

An SMB should be cautious when:

  • Its processes are not stable enough to encode.
  • The desired capability is genuinely generic.
  • The owner underestimates maintenance.
  • No internal person can own the system.
  • The project depends on one irreplaceable freelancer.
  • The company wants to preserve bad processes rather than improve them.
  • The business case relies only on eliminating subscription fees.
  • Failure would threaten the company, but the budget cannot support high-assurance engineering.
  • A configurable vertical product already solves 80–90% of the problem.
  • The custom system would recreate accounting, payments, identity, or other mature infrastructure without a special reason.

Custom software can turn organizational confusion into expensive, executable organizational confusion.

The likely AI-enabled opportunity

The most promising journeyman offer may therefore be neither “I will build any app” nor “I will replace all your SaaS.” It might be:

I will map your operation, identify where people are acting as glue between systems, create a company-controlled data and workflow layer, give each role one coherent interface, and maintain it under a predictable contract.

AI lowers the cost of building connectors, interfaces, migrations, tests, and documentation. The engineer’s valuable contribution is deciding:

  • Which SaaS products to keep
  • Which workflows to redesign
  • Which data should be authoritative
  • Which exceptions require humans
  • Which custom pieces create real advantage
  • How the whole system can remain secure and maintainable

That is a much stronger proposition than selling code. It sells the business a coherent operating system without asking it to become a software company.

Explore conversation