Your spreadsheet has tabs about itself
The 'master' tab. The 'do not edit' tab. The macros tab. The lookup tab. The history-of-fixes tab. The spreadsheet is the app. It's just a bad one.
Software shaped
like your business.
The off-the-shelf tool almost fits. But the workflow your team actually runs has six steps the tool doesn't know about, three integrations it can't make, and a UI your people fight every day. We build the version of the software that's actually shaped like your operation. Lightweight, owned, modifiable.
02 / Signals
Five signs your
operation needs an app.
Workflow Automation handles tasks. Agents handle conversations. Custom Applications handle the operations that need a real interface: a place your team logs in, sees what's going on, does the work, and moves on. Here's how you know you've crossed that line.
Your SaaS bill is bigger than the value
Three platforms at $400/mo each, none of which do what you actually need, all of which require a workaround. You've assembled an app out of overlapping subscriptions.
The training doc is longer than the workflow
Your team has a 14-page wiki article on how to use the system. Read carefully, the wiki article is documenting workarounds for the tool's defects.
Reports require humans
Every Monday someone exports four CSVs, joins them in Excel, formats them, emails them. The reporting need is daily. The reporting reality is weekly. The gap is a person.
You're hand-keying between systems
Two systems that need to share data don't. Someone retypes things between them. Errors creep in. Sync drifts. The work has become 'keep the systems honest.'
03 / What we ship
Six application shapes
we build most often.
We don't build greenfield SaaS. We build the internal tool your team needs but no off-the-shelf product makes. Some live alongside what you have; some replace a $400/mo subscription with something better.
/01
Operator Dashboard
One screen. Everything the person running the business actually looks at. Pipeline, pulse, problems. Pulled live from your systems. Designed for the person who runs the company, not the person who built the CRM.
An example
Manufacturing client replaced three SaaS dashboards with one custom screen. The CEO checks it on Sunday night and walks into Monday with a plan.
Typical inputs
- KPI definitions
- Source systems (any)
- Refresh cadence
- Who logs in + permissions
- Web app · responsive
- Live data, no overnight batch
- Drill-downs to source
- Permissions per role
/02
Internal Tool to Replace a Spreadsheet
That spreadsheet your operation runs on. The one with macros, with the 'don't touch' tab, the one that breaks when someone sorts wrong. We build the version of it that's a real app. Same logic, real database, real permissions.
An example
Real estate brokerage replaced a 4,200-row deal-tracking spreadsheet with a custom app. Deal cycle time fell 18% in 90 days.
Typical inputs
- The spreadsheet itself
- Who edits what
- Any external systems it talks to
- What's broken or scary today
- Web app, modeled on the spreadsheet
- Real database (no more 'don't sort')
- Audit history of every change
- Row-level permissions
/03
Client Portal
Where your clients log in and see their stuff: projects, invoices, documents, status. Branded as you. Reduces the 'where are we on this?' emails by 90%. Stops being a place where your team has to forward attachments back and forth.
An example
Services firm cut client-status emails by 87% after launching a portal. NPS rose 14 points the same quarter.
Typical inputs
- What clients need to see
- Brand assets + voice
- Existing client data + accounts
- Authentication preferences
- Branded web portal
- Mobile-responsive
- Document + file sharing
- Notifications + alerts
/04
Quoting + Proposal Tool
Bigger than the Workflow Automation quote engine. A real app: saved templates, version history, approval flows, e-signature, post-signature handoff. The quote tool you wish your CRM had.
An example
B2B services firm reduced quote-to-signed cycle from 11 days to 3, with a 22% increase in close rate.
Typical inputs
- Past quotes + templates
- Pricing logic + approval rules
- Sales team workflow
- Signature platform preference
- Web app
- Stripe / DocuSign integration
- CRM sync
- Templates + variant management
/05
Field / Mobile App
For the team that's not at a desk. Service techs, inspectors, drivers, on-site staff. Works offline. Captures what they need to capture. Syncs when there's signal. Replaces the clipboard + photo + text-message + voicemail chain that exists today.
An example
HVAC service company replaced paper job sheets + texted photos with a field app. Job close time fell from 4 days to 4 hours.
Typical inputs
- Field workflow today
- Devices the team uses
- Offline / sync requirements
- Back-office integration
- Mobile-web app (no app-store hassle)
- Native PWA for offline
- Photo + signature capture
- Back-office sync
/06
Lightweight SaaS / Customer-Facing App
When the thing you want isn't an internal tool, it's a thing you charge for. We build the MVP-shaped version. Real product, real users, real payments. The version that proves the market before you go raise a round.
An example
Solo founder validated a niche SaaS with a 12-week build. Went from idea to $4K MRR in 6 months, without hiring engineering.
Typical inputs
- The problem you’re solving + for whom
- Pricing hypothesis
- Auth + billing requirements
- What 'done enough to ship' means
- Production web app
- Stripe billing
- Multi-tenant from day 1
- Source code you own
04 / How it works
The five stages,
shaped for Custom Applications.
Every Sling engagement runs through the same five stages: Assess, Design, Build, Train, Optimize. Here's what that looks like specifically for Custom Applications. See the full Framework →
Map the operation
Half-day with the people who'll use it. We watch the current workflow, document the workarounds, identify what off-the-shelf almost covers and why it doesn't.
Spec + prototype
Solution Spec your team can read in 30 min. Working prototype of the riskiest screen. Tech stack, data model, hosting decisions, all on paper before code.
Ship weekly
Weekly working build, deployed to staging. Your team uses it on real workflow before it's done. Course corrections happen weekly, not at the end.
Roll out, document
Onboard the team. Document the system. Identify the in-house Champion who'll own it going forward. Run live sessions with real work.
Iterate against use
Ongoing changes billed by the hour, as you need them. The app meets real usage and reveals what we missed. We tune as it matures. Most apps settle within 4–6 months.
05 / Toolbox
What we typically reach for
on application builds.
Tool-agnostic by design. We pick what fits your data, your stack, and your budget — frontier models when they win, open-weight and on-prem when they're smarter. Nothing here is locked in. If you've already standardized on something else, we work with it.
Primary
Claude Code
How we build it. We don't outsource. We ship the code ourselves, in Claude Code, daily.
Primary
Claude (Anthropic)
The brain inside the app, when the app needs one: extraction, drafting, classification, summarization.
Framework
Next.js · React
The default web stack. Boring on purpose. Hires-able. Lots of community help if you ever move off us.
Backend
Convex · Supabase · Postgres
Where the data lives. We pick based on real-time needs, team comfort, and existing infrastructure.
Auth + payments
Clerk · WorkOS · Stripe
The plumbing. We don't build what these do well. We wire them up clean.
Hosting
Vercel · Cloudflare · AWS
Where it runs. We pick based on data residency, traffic, and your existing footprint.
06 / Proof
One shipped example.
A representative engagement. Anonymized industry tag, real-shape metrics. The full case-study library lives at /case-studies.
18% shorter deal cycle
Replaced a 4,200-row deal-tracking spreadsheet with a real app. Cycle time fell 18% in 90 days. The 'don't sort that' rule is no longer a rule.
The shape
- Assess: shadowed two agents through a full deal cycle. Mapped the spreadsheet's actual workflow vs. what was documented. Found 14 implicit rules nobody had written down.
- Design: data model + 6 core screens. Working prototype of the deal-detail page. Spec accepted in 2 rounds.
- Build: 6-week build. Team using v0.7 in week 3. Migrated 4,200 historical rows cleanly. Code in client's repo.
- Train: 2-hour onboarding for the 22 agents. Champion identified on the brokerage team. Documentation in their Notion.
- Optimize: ongoing changes billed by the hour. Added 4 features in months 2–4 based on agent feedback.
07 / Pricing
Simple.
Hourly. Honest.
The Discovery Call is free. Everything after, the audit, any foundations work, and the build, is billed by the hour. How many hours it takes depends on your size and scope, which we estimate up front. The rate is given on the Discovery Call, in writing. How pricing works →
Free
- Honest fit assessment
- A realistic sense of the hours
- No deck, no pitch
Hourly
- We scope the real work
- Fix foundations first if needed
- Hours estimated before we start
Hourly
- You see every hour
- Only pay for what we ship
- Source, data, credentials yours
08 / FAQ
Common questions
about this service.
What buyers actually ask on Discovery Calls about Custom Applications. If yours isn't here, we'll answer it on the call.
Use the off-the-shelf tool if it actually fits. We'll tell you on the Discovery Call. We build custom when the gap is real: when the workarounds are eating hours, the data is fragmented across subscriptions, or the team's workflow has shape no SaaS will ever cover. Sometimes the answer is 'switch SaaS,' not 'build custom.' We say so.
You do. Source code in your repo from day one. Documentation in your wiki. Credentials in your secret manager. We retain zero rights. If you fire us next week, you can hire any developer to pick it up. That's a feature, not a bug.
They will. They always do. Weekly builds mean you see what we're shipping before we ship the next thing. Small changes are absorbed; bigger ones get re-quoted transparently. We don't sneak change orders. We don't tell you something is impossible when we mean expensive. The conversation is straight.
Yes. We write boring, readable code in popular frameworks (React, Next.js, Postgres) on purpose. If you have a developer, even part-time, even fractional, they can pick it up. If you don't, we keep iterating with you, billed by the hour as you need it. Lock-in is the opposite of what we sell.
No-code is great for prototypes and admin panels. It hits a ceiling fast: performance, escape hatches, hiring. We build with real code, in stacks your future engineer can hire into. If no-code is actually the right answer for your problem, we'll say so and help you pick the right tool. We've sent clients to Retool. We will again.
09 / Next stepShow us the
Show us the
spreadsheet that
runs your operation.
The one with macros, with the 'do not edit' tab, with the workarounds. Bring it to the Discovery Call. If a real app is the right shape, you'll leave with a realistic estimate of the hours, all billed by the hour. If switching SaaS is the better play, we'll say that too.
What this call is
30 min. We learn your operation, scope the work, tell you straight whether Custom Applications fits.
What you'll leave with
A realistic sense of the hours and the scope, or an honest "this isn't us" and a referral to a better fit.
Other services
If the work is conversational, see AI Agents. If it's task automation without UI, see Workflow Automation. Most apps are a combination of all three.