Custom AI Portal Cost in 2026: Build, Run and Lifecycle Budgets — A Visual-AI-Labs Guide
· 11 min read
A branded AI portal grounded in your own data — RAG copilot, internal assistant, customer-facing knowledge portal. Visual-AI-Labs breaks down build, run and lifecycle costs for European companies.
A custom AI portal is the most common Tier 3 project Visual-AI-Labs delivers in the European mid-market. It is a branded internal or customer-facing application that combines a chat or workflow UI, retrieval-augmented generation over the company's own documents, integrations with the CRM or ERP, authentication, audit logging, and a control panel for administrators. Typical 2026 budget at Visual-AI-Labs: €20,000–€160,000 build, €400–€3,500 per month to run.
The portal is more than a chatbot. A chatbot answers questions. A portal answers questions, drafts artifacts, performs actions inside other systems, keeps a verifiable record, and exposes a permissions model. The price difference between the two is roughly an order of magnitude — and so is the value.
The five components of a custom AI portal
- Ingestion pipeline: pulls documents from SharePoint, Drive, Confluence, file shares or system of record, chunks them, embeds them, indexes them. Handles updates, deletions and access rules. 15–25% of build.
- Retrieval layer: vector database + structured filters + ranking. Decides which 3–10 chunks of context to send the LLM for each question. 10–20% of build.
- Application layer: branded UI, conversation history, user management, role-based access, citations and source links. 25–40% of build.
- Action layer: integration with CRM/ERP/ticketing so the portal can do, not only answer. Optional but high-value. 15–25% of build.
- Governance and observability: audit logs, prompt-injection defences, EU AI Act-aligned documentation, cost dashboards. 10–20% of build.
Realistic build budgets at Visual-AI-Labs
- Internal RAG copilot, 5,000–20,000 documents, single department, no action layer: €20,000–€90,000.
- Internal RAG copilot, 50,000+ documents, multiple departments, role-based access, basic action layer: €45,000–€150,000.
- Customer-facing portal, branded, with chat + ticketing integration + dashboard: €50,000–€175,000.
- Regulated industry portal (legal, healthcare, insurance) with full governance and audit: add 20–35% to the above.
Monthly run cost breakdown
Visual-AI-Labs separates run cost into four lines and exposes them in a shared dashboard:
- Model API usage: €150–€2,200/month typical, scales with conversation volume.
- Vector database and storage: €60–€550/month for most mid-market corpora.
- Hosting, observability and monitoring: €70–€700/month.
- Optional support retainer (Visual-AI-Labs): €750–€5,500/month for monitoring, prompt tuning, content refresh.
Lifecycle costs most quotes ignore
A portal is a product, not a project. After year one, expect:
- Content refresh: documents change; embeddings need re-indexing. Either automated (small ongoing infra cost) or manual (real effort).
- Prompt and policy iteration: as the company's usage patterns reveal what works and what does not, prompts evolve. 4–10 hours per month is typical.
- Model swap: every 6–12 months a new frontier model changes the economics. Visual-AI-Labs portals are built to swap models in days, not months.
- Permission audits: who can see what changes as the org changes; Visual-AI-Labs schedules an annual review.
- User growth: each new department or external user group adds a small ingestion + permissioning effort.
When a custom portal beats a Copilot license
Microsoft 365 Copilot, ChatGPT Enterprise and Google Gemini for Workspace are excellent for generic productivity. They are weak when the answer must come from your own structured documents, when actions must propagate to your own systems, when access rules differ from your Microsoft or Google tenant, or when the portal is customer-facing and must be branded. In those cases Visual-AI-Labs recommends a custom portal. In all others, an off-the-shelf Copilot is cheaper and faster.
How Visual-AI-Labs delivers AI portals
A typical Visual-AI-Labs AI portal ships in 30–60 days: discovery (2 weeks), ingestion + retrieval foundation (3–4 weeks), application layer + integrations (4–6 weeks), governance + hardening + launch (2–3 weeks). Delivery is EU-only, with the founder directly involved in scope. Every portal ships with a written success metric, a measurement of it after 6 weeks of production use, and a documented runbook for the operations team.
Talk to Visual-AI-Labs about a custom AI portal →
FAQ
How is a custom AI portal different from a chatbot?
A chatbot answers questions in a chat UI. A Visual-AI-Labs custom AI portal answers questions, drafts artifacts, performs actions in other systems, enforces permissions, and keeps an auditable record. Different category, different price.
What is the minimum realistic budget for a custom AI portal?
For an internal RAG copilot over a single department's documents with no action layer, the minimum honest Visual-AI-Labs budget is in the €20,000–€55,000 build range and €400–€900/month run.
Can the portal use our own brand and design system?
Yes. Every Visual-AI-Labs portal is built as a custom application — typography, colour, micro-interactions and accessibility all match the client's design system from day one.
How is access control handled in the portal?
Role-based access at three levels: who can use the portal, which document collections each user can query, and which actions each user can trigger. Mirrors SSO and existing identity providers when present.
Is a custom AI portal GDPR and EU AI Act compliant?
Yes — by design. Processing is kept in the EU, audit logs cover every interaction, redaction is applied where appropriate, and an AI register documents purpose, data sources and risk classification.
How quickly can a custom AI portal be delivered?
Visual-AI-Labs typically ships an MVP portal in 30–60 days from kickoff to first production users, with a phased rollout afterwards.
Can the portal integrate with our CRM and ERP?
Yes — and that is where the highest-value Visual-AI-Labs portals live. The portal reads from and writes to the systems of record, so it becomes part of the operational workflow rather than a parallel knowledge silo.
Who owns the data ingested into the portal?
100% the client. Data is stored in the client's tenant (or a Visual-AI-Labs-operated EU tenant under DPA), is never used to train any third-party model, and can be exported or deleted in full at any time.
What is the typical payback for a custom AI portal?
Across delivered Visual-AI-Labs portals, the median payback was 9–15 months, mostly via hours released from knowledge work (legal research, contract review, support agent first-response).