Website, Product, or Internal Tool: What You Need Next

The 30-second answer

Build a website when your buyers can’t tell what you sell, your brand doesn’t match the product behind it, or you’re leaking traffic from a homepage that doesn’t convert.

Build a product when the thing your customers pay for is something they use, not something they read. A platform, an app, a tool. The website is the shop window. The product is the shop.

Build an internal tool or automation when your team is losing 5+ hours a week per person to manual work, copy-paste between systems, or processes that only one person understands.

Most businesses need one of these urgently and the others later. Pick the one that is actively blocking growth right now. The other two will still be there in six months, and they’ll usually be cheaper to build once the first is working.

The three options, side by side

When a website is the right next spend

The most common tell: your product or service has moved on, and your website hasn’t caught up. You’re selling a 2026 offer through a 2022 site.

Signs it’s the right call:

• Homepage bounce rate above 70%

• Your team spends the first 20 minutes of every sales call explaining what you do

• SEO is flatlining and AI search engines don’t cite you

• The site was built by a founder in a weekend, or an agency that ghosted

• You have a launch, a rebrand, or a repositioning coming

• Your best salespeople keep sending prospects Loom videos instead of your homepage

A good marketing website in 2026 is not decoration. It is the highest-leverage sales asset most B2B businesses own. It works while you sleep, filters bad-fit leads, and shortens the sales cycle for the good ones. If we had to pick one asset a £2M–£25M B2B should get right first, it is almost always this.

Real client example. Reka, an AI foundation model company backed by NVIDIA, launches new models on the site we built. They needed pages that could ship fast, hit 100% Lighthouse scores, and read as premium enough to sit next to Anthropic and OpenAI. Framer was the right platform. Their marketing lead ships new pages without waiting on a developer. That last bit matters. A site that only your dev team can update is not a marketing asset. It’s a bottleneck.

Another one. Sydecar, a US fintech on Series A capital, had a 50+ page site that had grown organically for years. Nobody owned it. We rebuilt it as a system so their growth marketer could run experiments and add pages without breaking anything. Post-launch, HubSpot attribution actually works.

The pattern is the same. A lean marketing leader who owns the whole surface. No internal web team. The site was slowing them down. If that sounds familiar, a well-designed Framer marketing site is probably the single biggest lever you have.

Framer vs Webflow vs custom code

Nine out of ten B2B marketing sites we build now sit on Framer. It gives marketers real editing power, hits 100 Lighthouse scores without special effort, and the CMS is fast enough for weekly content shipping. Webflow made sense in 2020. In 2026 it’s mostly inertia. Custom Next.js only wins when you’re integrating tightly with a product, running a huge blog with complex taxonomies, or your team already lives in React. If you’re weighing that decision, our full breakdown in Framer vs Webflow vs custom for marketing-led SMBs walks through the real tradeoffs.

What it costs

• Discovery Sprint: £8k–£15k, 2–3 weeks, money-back guarantee. Wireframes, brand direction, information architecture, and a locked scope for the build.

• Framer marketing site build: £15k–£40k, 4–8 weeks. Includes design, build, CMS setup, and a handover your marketer can actually use.

• SEO and content retainer: from £3k/month if you want us to keep it moving. Optional.

If someone quotes you £60k+ for a marketing site, ask what you’re getting that a top-tier Framer build doesn’t give you. Most of the time, the answer is a slower CMS and a bigger dev bill next year.

When a product build is the right next spend

You need a product when the thing you sell is something people log into. A platform. An app. A marketplace. A workflow tool. The website is the front door, but the product is the business.

Signs it’s the right call:

• You have paying customers using a prototype that keeps breaking

• You’re on Bubble or Lovable and hitting the ceiling

• You’ve raised money and told investors you’d ship something

• A manual process (Airtable, Google Sheets, an ops person) is the product today, and you need to productise it

• Your CTO left and the codebase is a black box

• Your competitors are shipping AI features you can’t match without proper infrastructure

This is the biggest spend of the three. It should also produce the biggest return. A working product is a defensible asset. A marketing site is not. When it works, it becomes the balance sheet: recurring revenue, customer stickiness, and a moat competitors can’t paint over.

Real client example. Just Video Walls came to us with a manual quoting process for enterprise video-wall installations. We built them a Bubble configurator with seven user roles, connected to HubSpot, replacing weeks of back-and-forth per quote. The build paid back inside two quarters through faster sales cycles and fewer errors.

Another one. Astrid, a founder-led AI product, was shipping features on Lovable but had no way to deploy safely. We took the vibe-coded prototype, wrapped it in GitHub Actions, Supabase migrations, and a proper Vercel CI/CD pipeline. Fixed five hidden migration bugs on the way. They can now ship changes without praying.

Third one. MeltX runs an FCA-regulated, SEC-registered institutional finance auction platform. Over $2 billion has been listed on it. We built it, we still run it, and it has zero major incidents. This is what a product build looks like when the stakes are real. Regulated finance is not the place to test whether your dev shop knows what they’re doing.

MVP or production-ready, honestly

There’s an important sub-question inside the product decision: are you validating something, or scaling something? MVPs on Bubble are the right answer if you need to test whether anyone will pay for the thing at all. They ship fast, cost less, and scale to your first few hundred users comfortably. Production-ready Next.js builds are the right answer if you already have paying customers, or if your first users need audit trails, SSO, and 99.9% uptime.

The mistake we see most often: founders picking Bubble because it’s cheap, trying to run 10,000 concurrent users through it, then blaming the tool. If you’re past product-market fit, don’t cheap out on the stack. If you haven’t found it yet, don’t over-engineer.

We’ve written a longer piece on this exact fork: Bubble to real code: when to migrate (and when not to). Read it before you write the RFP.

What it costs

• Discovery Sprint (highly recommended): £8k–£15k. Non-negotiable for anything above £40k. It de-risks the whole engagement.

• MVP / Sprint build: £40k–£80k, 8–12 weeks. Bubble or Next.js. Enough to validate, sign customers, hit a milestone.

• Production-ready build: £80k–£150k+, 12–20 weeks. Custom stack (Next.js, Postgres, Vercel, Clerk or Auth0). What you want if you have paying customers on day one.

• Post-launch retainer: Maintain tier from £3k/month, Grow tier from £8k/month.

We price in two options, never three. Sprint vs Premium (or MVP vs Production-Ready). Same rule for everyone. If you’d like to see how a serious custom SaaS build is scoped, our services page walks through the full lifecycle.

When an internal tool or automation is the right next spend

This one is the most underrated. It doesn’t show up on your homepage, so leadership rarely prioritises it. But if your team is copy-pasting between HubSpot and a spreadsheet, chasing suppliers by email, or rebuilding the same report every Monday, you are paying for it in hours you can’t get back.

Signs it’s the right call:

• Any single workflow eats more than 5 hours a week across your team

• You’ve hired to keep up with manual work rather than growth

• Two systems don’t talk to each other and someone in the middle is the integration

• You want to add AI to ops but don’t know where to start

• The board is asking about AI ROI

Internal tools and n8n automations have the fastest payback of the three. Often inside a quarter. Sometimes inside a month. If your COO or ops director is losing weekends to reporting, this is where you should be spending.

Real client example. HubSync, a US fintech, had ~23,000 contacts split across HubSpot and their own platform. We built a two-way n8n sync, plus a newsletter workflow, a podcast production automation, and a Sales Navigator prospecting flow. Their marketing lead runs the whole GTM engine solo now. That is what “AI transformation” actually means for a lean team. Not a McKinsey deck. A working stack that lets one person do the work of five.

Another one. Stratverse, a consulting marketplace, had six manual ops eating their week. We automated all six. They got back more than 40 hours a week. That is a full extra person, without the hiring cost.

Third one. Pockla was pre-seed when we built their MVP with n8n and Slack automations. That build helped them close a £1.6M seed round. Automation as a fundraising asset. Not what most people expect.

Do it in-house or hire an agency?

There’s a genuine question about whether your ops director should just learn n8n themselves. For simple triggers between two SaaS tools, they probably should. For anything involving state, retries, error handling, or a database, the DIY approach ends in tears within six weeks. We wrote a full decision framework in Should your ops director learn n8n or hire an agency?, read it before you make the call, especially if you’re a £5M+ business.

If you’re specifically scoping n8n work, the UK market has thinned out considerably in the last year. Our best n8n agencies in the UK guide covers who’s actually shipping.

What it costs

• Discovery Sprint (optional here): £8k–£15k if the scope is fuzzy.

• Single automation or internal tool: £8k–£20k, 3–5 weeks.

• Ops automation retainer: from £4k/month, ongoing.

• Complex internal platform: £20k–£60k, 6–10 weeks.

If a payback is more than six months on an internal build, we’ll tell you. The whole point is fast ROI. For a fuller price breakdown across sprint types, see our AI automation sprint cost guide for 2026.

The one nobody asks about: when the answer is “all three, sequenced”

We built this piece to help you pick one. But there’s a common pattern for growing B2B companies where the honest answer is that you need all three, in a specific order, across a year. If that’s you, here’s the sequence we usually recommend.

Quarter 1: Website. Fix positioning first. If your buyers can’t tell what you do, nothing else you build matters. £20k–£40k, 4–8 weeks. A Framer site with a proper CMS and analytics setup.

Quarter 2: Automation. Now the marketing engine has a front door, plug in the pipes. HubSpot cleanup, lead-routing, MQL scoring, a Slack digest for sales. £10k–£25k, 3–5 weeks. ROI stacks fast here, because every good lead now flows without human effort.

Quarter 3–4: Product or internal platform. With the marketing and ops layer solid, you build the thing that actually differentiates you. £60k–£150k, 12–20 weeks.

Sequenced this way, each build feeds the next. The site pays for the automation through better inbound. The automation pays for the product through operational leverage. By the time the product ships, you have a marketing engine and an ops backbone waiting for it. That’s what we mean by sequencing rather than three parallel projects your team can’t absorb. Our case studies page shows how this played out for HubSync, Sydecar, and MeltX.

Where each one breaks

None of these are magic. Being honest about the failure modes is the difference between a happy client and an angry one.

Websites break when you treat them as a one-off. A site launched in Q1 that nobody touches is dead weight by Q4. Budget for either an internal owner or a retainer. Otherwise you’ll be having this conversation again in 18 months.

Product builds break when the founder wants the entire wishlist in v1. Every product we’ve ever shipped that worked was ruthlessly cut down before launch. If you can’t say what the MVP does in one sentence, you’re not ready. The Discovery Sprint exists for this reason.

Internal tools break when nobody adopts them. The best n8n automation in the world is worthless if the team routes around it. You need executive air cover and a lead user before we start. This is why we ask, on every discovery, who owns adoption. If the answer is “nobody yet,” we pause.

The AI question everyone’s asking

Every one of these three surfaces has an AI angle in 2026. Which is why “we need AI” is not a useful brief.

On the website: AI is now a distribution channel. If Perplexity, Claude, and ChatGPT can’t summarise your homepage cleanly, you’re invisible to a growing share of buyers. Structured content, clear positioning, and machine-readable metadata matter more than they did last year.

On the product: AI is a feature. It’s rarely the differentiator on its own. The winners are products that use AI to remove a specific painful step for the user, not products with a chatbot bolted onto the sidebar.

On internal tools: AI is genuinely transformational. The step-change is here. Claude Code and similar agentic tools let us ship internal software 3–5x faster than 2023. That’s a real cost curve shift, and it changes the calculus on which internal builds are worth doing at all. If you want a full view of who’s actually shipping this way, our take on UK Claude Code agencies in 2026 covers who’s real and who’s mostly LinkedIn.

And if the AI conversation at your board has escalated into “should we hire McKinsey or Accenture,” that’s a separate question with a lot of expensive wrong answers. We covered the honest UK-agency alternatives in AI transformation firms: UK alternatives to McKinsey and Accenture.

How to actually decide

Ask yourself three questions.

1. Where is revenue actually leaking? Bad-fit leads (website), poor product experience (product), or ops overhead (internal). Whichever bleeds most is where you start. Don’t guess. Look at the numbers.

2. What can your team maintain? A £40k marketing site with no one to update it is a bad investment. A product with no PM is worse. Be honest about who owns the thing after we ship. If the answer is “we’ll figure it out,” you’re not ready.

3. What does the next 12 months look like? Fundraise? Product launch? New market? Some builds only make sense in specific windows. Don’t build a website six weeks before a rebrand. Don’t ship a product v1 the same quarter you’re moving your CTO.

If you’re stuck between two, do the Discovery Sprint. £8k–£15k, 2–3 weeks, money back if you don’t like what we produce. You leave with a locked scope, a real budget, and the confidence to green-light the build. Or the confidence to walk away.