Customer and client portals, web applications, iOS and Android apps, websites, internal tools and line-of-business systems. We build the thing that does not exist as a product at a price you can afford — and the thing your team is currently doing by hand, in a spreadsheet, twice.
In a survey of 300 operators who had commissioned a custom build, the reasons clustered into three. Only one of them is the one vendors like to lead with, and it is the weakest of the three.
A third built because the market did not serve them. The product that does the job is enterprise-priced, or it is aimed at a business ten times their size, or the thing they need sits between two categories and nobody sells the middle. This is the strongest case for building, and the least contested.
A fifth replaced a manual process — spreadsheets emailed around, paper forms retyped, the same order entered into two systems, a person whose job is partly to be the integration between two products. That labour is a real, ongoing cost, and it is usually the easiest one to quantify honestly.
A third dropped a SaaS subscription. Rarely because it was expensive on its own — usually because the fit gap had grown into a second, unofficial system of workarounds, and they were paying for the product and paying for the workaround.
No per-seat licence, no renewal negotiation, no feature you need sitting on someone else's roadmap, no repricing when you grow. That is real and it matters — but it is what closes the decision, not what should start it. If ownership is the only argument, the numbers probably do not support the build.
The single most common request we get. A customer portal is a private, logged-in place where the people you serve can do for themselves what they currently do by phoning your office — and where your team stops retyping the same information into three systems.
Accounts and roles, job or order status, documents and invoices, uploads and approvals, payments, messaging, notifications, and a reporting view for the account owner. Which of those you actually need is the first conversation, and it is usually fewer than the list above.
It is permissions and data. Who can see what, at which organisation, in which role — and where the real record lives, so the portal reads from your existing system instead of becoming a fourth copy of the truth that quietly drifts out of date.
They model a business that is not yours. The status values are wrong, the hierarchy is wrong, the approval step you need does not exist, and the workaround becomes a spreadsheet nobody admits to. If you can live inside the product's model, buy it. If you have already stopped being able to, build a customer portal that matches how your work runs.
Same architecture, different emphasis: engagements rather than orders, document exchange, review and sign-off, and a clean separation between what a client sees and what stays internal. Client portal development usually starts smaller than customer-facing work and grows once people trust it.
Custom content across the board — AR, VR, web, mobile, WebXR, websites, or rebuilding a system you have outgrown. Software is not a side line here; it is the same team, the same process.
Logged-in areas for the people you serve, connected to the systems that already hold your data. Roles, status, documents, approvals, messaging.
Real applications in the browser, not brochure sites with a form. Dashboards, scheduling, work orders, quoting, data capture, reporting.
Native iOS, or one cross-platform build serving both. Offline mode, camera and photo capture, geolocation, push — priced as add-ons so you can see what each one costs before you commit.
The systems your company runs on. We have built a medical transcription platform, scoped a work-order platform for a flooring contractor, and shipped an AI roleplay coaching app — this is genuine work, not a category we added to a list.
Marketing sites that load fast, rank, and are editable by the people who need to edit them. Often the front door to the portal, built at the same time.
A chatbot trained on your own knowledge base, a coaching or roleplay experience, or an agent that takes multi-step action in your systems. Built into the product rather than bolted on as a separate tool nobody opens.
If a vendor opens with "stop paying subscriptions and save money", check their arithmetic. Over three years the totals usually land close together: one published model puts a custom build at roughly $84,000 against roughly $89,000 for SaaS at 25 seats. Maintenance is the reason — budget around 20% of the build cost per year, every year. A buyer who runs the numbers will find this, so we would rather hand it to you first.
Seat counts high enough that per-user pricing compounds against you. Work being done manually because the product cannot do it. A capability that is core to how you compete and that you do not want to rent. Those three carry a build. "It feels expensive" does not.
Commodity problems with mature products — accounting, payroll, email, ticketing, checkout. Anything where compliance is the hard part and the vendor carries it. Anything where the product is 80% right and the gap is a preference. We say this on first calls and we lose work to it.
Vertical software is good at the workflows it was designed around, and plenty of companies should stay on it. We make no claims about anyone's pricing, contract terms or exit costs — those are not published, and every figure in circulation is unverified. Read your own agreement. Then look at what you actually use.
Almost nobody uses most of the platform they pay for. The realistic path is scoping the 20% your business runs on, building that properly, integrating it with what stays, and migrating in phases — not a big-bang replacement of a system that took a vendor a decade to write.
We use AI heavily in delivery and it is genuinely faster to a working version. What it has not changed is everything after that, which is where projects are won or lost.
A 2025 MIT study of enterprise AI, circulated as The GenAI Divide out of the Project NANDA initiative, found roughly 95% of enterprise generative-AI pilots produced no measurable P&L impact — and that externally-sourced tools reached production roughly twice as often as internally built ones. The authors labelled these preliminary findings, so read it as directional rather than settled.
A prototype in a week is not the same as a system your customers log into. Someone has to read the code, understand the edge cases, and own the decisions the model made silently. We do that; the speed shows up as lower cost, not as skipped steps.
Auth, permissions, data handling, dependency hygiene and the boring parts of running something on the public internet are unchanged. If anything, faster generation makes it easier to ship a plausible-looking system with real holes in it.
Roughly 20% of build cost per year, in our experience and in the published models. We quote it up front rather than pretending it does not exist, because it is the number that decides whether custom was the right call three years from now.
We map the workflow you actually run, not the one the org chart describes, and separate what must be built from what can be bought or deferred. For AI work this is a fixed-price readiness audit — $4,500, two weeks, and you keep the findings whether or not you build with us.
A prototype or interactive spec before production code, so the flow gets argued about while changes are still cheap. This is where the rework is supposed to happen, and where most fixed-price projects go wrong when it is skipped.
Working software in front of you at intervals rather than a reveal at the end. We quote in ranges and phases rather than inventing a delivery date — for reference, our measured immersive builds run about twelve weeks for a full Unity module and three to four weeks for a 360 module, and software of comparable scope behaves similarly.
One structured round of user acceptance testing, then deployment, documentation and the source code. Additional scope is quoted separately rather than absorbed silently, and maintenance is priced in writing before you sign, not discovered afterwards.
We publish our prices. Almost nobody in this industry does, and it is the single most common complaint buyers have about it.
| Scope | Typical use | Price |
|---|---|---|
| Quick interactive experience | One workflow, a calculator, a configurator, a focused internal tool | $5,000 – $10,000 |
| Interactive platform | Customer or client portal — accounts, roles, multiple modules, status tracking | $10,000 – $15,000 |
| iOS app | Native iPhone and iPad, single platform | $20,000 |
| Cross-platform app | iOS and Android from one codebase | $30,000 |
| Chatbot on your knowledge base | Support deflection, internal Q&A over your own documents | $6,000 – $12,000 |
| AI coach or roleplay app | Scored conversational practice with feedback | $8,000 – $18,000 |
| Custom AI agent | Multi-step, takes action in your systems | $15,000+ |
| AI readiness audit | Two weeks, fixed price, before you commit to a build | $4,500 |
| AI pilot sprint | 4–6 weeks, one workflow, running in production | from $15,000 |
| AI partner retainer | Ongoing build, maintenance and iteration | from $5,000/mo |
Common add-ons: additional screens +$2,000 · in-app purchases +$4,000 · offline mode +$5,000 · social login +$2,500 · camera and photo capture +$3,000 · geolocation and maps +$3,500 · real-time chat +$6,000 · video streaming +$7,000 · advanced analytics +$2,500 · knowledge base or RAG +$3,000 · voice interface +$4,000 · custom branding +$1,500 · tool and API integrations quoted per system. Terms are 30% at kickoff, 35% at UAT and 35% on final delivery, with one round of UAT included and additional scope quoted separately.
Industrial, energy, utility, healthcare and food-service operators — plus the software companies and contractors who need a product built rather than a pilot run.
A medical transcription platform — the product itself, designed and built as working software rather than a demo.
A work-order platform scoped and specified around how the crews actually run jobs, rather than around a product's idea of a job.
Conversational practice with objectives tracked and a scored result at the end — speech in, speech out, built as an application.
A browser-delivered site walk with an AI companion, built bilingually in English and Spanish for a construction environment.
Customer portal development builds a private, logged-in area where your customers or clients do the things they currently do by email and phone — check the status of a job or order, download documents and invoices, upload what you need from them, approve something, message the person handling their account. A client portal is the same idea for professional services. The work is rarely the login screen. It is deciding what each role can see, and connecting the portal to the systems that already hold the real data, so nobody has to keep two records in sync by hand.
We publish our prices. A focused single-workflow tool runs $5,000 to $10,000. An interactive platform — accounts, roles, multiple modules, progress and status tracking, which is what most customer web portals actually are — runs $10,000 to $15,000. A native iOS app is $20,000 and a cross-platform iOS and Android build is $30,000. Add-ons are priced separately: social login +$2,500, real-time chat +$6,000, offline mode +$5,000, advanced analytics +$2,500, and so on. Terms are 30% at kickoff, 35% at UAT and 35% on final delivery.
Often not, on pure cost. Over three years the totals tend to land close together: one published model puts a custom build at roughly $84,000 against roughly $89,000 for SaaS at 25 seats. Maintenance is why — budget somewhere around 20% of the build cost every year, and anyone who tells you otherwise is quoting you half a project. Custom makes financial sense when the seat count is large enough that per-user licensing compounds, when no product on the market does the job at a price you can afford, or when people are absorbing the gap manually and that labour is the real line item. It does not make sense as a way to shave a subscription off next quarter.
Buy when the problem is genuinely common and well served — accounting, payroll, email, CRM basics, helpdesk ticketing, e-commerce checkout. Buy when compliance is the hard part and a vendor carries that burden for you. Buy when the product is 80% right and the missing 20% is a preference rather than a cost. We would rather tell you that in the first call than sell you a build you did not need, and we do tell people that.
Sometimes, and it depends on what is actually wrong. Vertical products like these are good at the workflows they were designed around. People come to us when the fit gap has started costing real money — the way their business runs does not map to the product's model, the module they need is on someone else's roadmap, or their team has built a shadow system of spreadsheets around the software they pay for. We do not make claims about any vendor's pricing, contract terms or exit costs; those are not published, and you should read your own agreement. What we will do is scope the subset you actually use, price it, and be straight with you if the honest answer is to stay where you are.
You own it. The application, the client-specific code and the data are yours, with no per-seat licence, no renewal negotiation and no feature you need sitting on a vendor roadmap. We retain only our general-purpose tooling, which you get a licence to use as embedded in the deliverable. Maintenance is quoted rather than ignored — hosting, dependency and security updates, platform changes like a new iOS release, and the small changes every live system needs. Plan for roughly 20% of the build cost per year, or work with us on a partner retainer from $5,000 per month.
Some teams can, and if you have engineers with capacity you should try. The best available evidence says it is harder than it looks: a 2025 MIT study of enterprise AI, circulated as The GenAI Divide out of the Project NANDA initiative, found roughly 95% of enterprise generative-AI pilots produced no measurable P&L impact, and that externally-sourced tools reached production roughly twice as often as internally built ones. The authors labelled those preliminary findings, so treat it as directional rather than settled. It matches what we see: AI gets you a working prototype fast, and then review, security, integration and maintenance are the parts that decide whether it ever goes live.
Send us the spreadsheet, the workaround, or the product you have outgrown. We will tell you what it would cost to build, what it would cost to keep, and whether you should buy something instead.
Start a conversation