Custom AI Chatbot vs Off-the-Shelf: An Honest Build-vs-Buy Framework

Every chatbot vendor will tell you to buy. Every AI agency will tell you to build. Both are answering the question "how do I get paid?" rather than the one you asked.

We build custom chatbots for a living, so read this knowing our bias. But the framework below is the one we actually use in discovery calls, and it regularly ends with us telling a prospect to go buy an off-the-shelf tool. Here it is.

Question 1: How standard is your workflow?

Off-the-shelf chatbot platforms are excellent at the median use case: FAQ answering over a knowledge base, order-status lookups, ticket deflection with handoff to a helpdesk. If your needs match the median, a good SaaS bot deployed in a week beats a custom build finished in two months.

Custom earns its cost when your workflow is structurally unusual:

• Answers depend on combining your internal systems (entitlements, account state, inventory) with policy, not just documents.
• The bot must take actions: issue refunds, modify bookings, file cases, under rules specific to your business.
• You need behavior that platforms won't let you control: custom escalation logic, domain-specific safety rules, unusual languages or channels.

Rule of thumb: if you can describe your bot's job in one sentence a platform's sales page already uses, buy. If your sentence has three "except when" clauses, build.

Question 2: How deep do integrations go?

Integration depth is where "configurable" platforms quietly become consulting projects billed at platform prices.

• Shallow (docs plus one helpdesk): platforms shine, buy.
• Medium (CRM lookups, order systems, auth): possible on platforms, but you'll live inside their connector limits. Compare that cost, including the workarounds, against custom before deciding.
• Deep (multiple internal systems, permission-aware answers, actions with approval flows): custom is usually cheaper over 18 months, because you're not fighting an abstraction that wasn't designed for you.

A detail that bites later: permission-aware retrieval. If different users may see different documents, the bot must respect source-level permissions at answer time. Many platforms handle this poorly or not at all, and retrofitting it is miserable.

Question 3: How sensitive is the data?

If conversations will carry customer PII, financial records or health data, ask any platform vendor three questions: where is data processed, who can see transcripts, and what's used for training? The answers are sometimes fine, and sometimes disqualifying, especially with data-residency requirements.

A custom build lets you choose the model provider per your constraints (including deployments where nothing leaves your cloud), log what you decide to log, and put your compliance story in writing. That control is a real cost driver toward custom for regulated industries, and irrelevant for a bot that answers shipping-policy questions. Be honest about which one you are.

Question 4: Does the volume justify engineering?

Rough shapes, which you should verify against your own numbers because pricing shifts constantly:

• Platform: per-seat or per-resolution pricing that's trivial at low volume and grows linearly forever. At high volume, per-resolution fees can exceed what the underlying model calls cost by an order of magnitude.
• Custom: an up-front build cost, then hosting plus model usage, which at today's model prices can be startlingly low. The margin platforms charge is the thing you're buying out of.

For scale on the build side: 2026 industry benchmarks put LLM-powered custom chatbots at roughly 8 to 25 lakh rupees from Indian engineering teams, while US and UK agencies quote 15,000 to 80,000 dollars for equivalent scope. That gap is worth knowing when a platform's annual bill starts approaching a build's one-time cost.

The crossover typically arrives with scale. If you're handling a few hundred conversations a month, the platform's margin is cheaper than an engineering project. At tens of thousands, the math usually flips, and the custom bot is also simply better, because it was built for your workflow.

The hybrid path most companies should consider

Build-vs-buy isn't one decision forever:

1. Start on a platform for the standard 70% of queries. Ship in weeks. Learn.
2. Instrument everything. The transcript stream tells you exactly where the platform fails: which questions go unanswered, which handoffs frustrate users, which workflows people keep asking for.
3. Graduate to custom when the failure list is long, specific and expensive. At that point you're building against evidence, not guesses, and the evidence makes the custom build cheaper and better-targeted too.

The mistake is staying on the platform for two extra years out of inertia while paying per-resolution fees on workflows it fumbles, or building custom on day one for a use case a 15,000-rupee-a-month tool handles.

Red flags in either direction

When buying, walk away from vendors who can't demo your documents live, who won't show groundedness or accuracy numbers, or whose citations are decorative rather than real sources.

When building, walk away from agencies that don't mention evals in the first conversation, can't explain their grounding and guardrail approach concretely, or quote a price before understanding your integrations. And if they promise "no hallucinations, guaranteed": run.

The checklist

Buy when: standard workflow · shallow integrations · low-to-moderate volume · no unusual compliance needs · you need it live this month.

Build when: the bot must act, not just answer · integrations are deep or permission-aware · data constraints rule out the platforms · volume makes per-resolution pricing silly · the bot is part of your product.

Still unsure after the four questions? That's genuinely common; the honest cases are close calls. A short consulting engagement that scores your specific situation costs a fraction of either mistake.

---

Mindela builds retrieval-grounded, brand-safe chatbots with evals and guardrails as standard, and tells prospects to buy off-the-shelf when that's the right answer. Ask us which side of the line you're on.