Zendesk no longer bills its AI Agents per license but per outcome: market analyses put the cost at an effective one to two US dollars per automatically resolved ticket (Richpanel analysis, 2026). That means your software vendor earns money on exactly the request volume a good customer portal is supposed to reduce for good. That's legitimate — but it shows why make or buy for a customer portal isn't a feature question. It's a question of how costs develop over five years.
We build customer portals for a living — and we still regularly advise clients to buy. The honest answer doesn't hinge on feature lists but on five criteria you can check cleanly for your own company. This article delivers the list prices on the buy side (as of July 2026), a decision framework, a worked 5-year cost example, and the third path that wins most often in practice.
The buy side: what portal software costs on paper in 2026
Zendesk publishes its prices transparently: according to the Zendesk pricing page (as of July 2026, billed annually), Support Team costs €19, Suite Team €55, and Suite Professional €115 per agent per month; Suite Enterprise is quote-only. The add-ons are where it gets interesting: Copilot adds another €50 per agent per month, Workforce Engagement another €50, and the Contact Center €82. If you price out a realistic full build-out — Suite Professional plus Copilot, Workforce Engagement, and Contact Center — Richpanel puts you at over 265 US dollars per agent per month (US list prices).
Freshdesk is the cheaper flavor of the same model: according to the Freshworks pricing page (as of July 2026, billed annually), Growth costs $19, Pro $55, and Enterprise $89 per agent per month; the Freddy AI Copilot is available from Pro upward for $29 per agent, and additional AI agent sessions cost $49 per 100 sessions. A plus for B2B scenarios: Pro includes 5,000 external collaborators.
And the enterprise class? Liferay DXP publishes no list prices — the subscription is calculated by environments and support tier. Third-party sources put enterprise subscriptions in 2026 at typically 25,000 to 50,000 US dollars per year and up; that's a market-observation estimate, not an official price. Anyone shopping in that league negotiates individually anyway — and should be running the make case in parallel at the latest by then.
As of July 2026. All SaaS prices are list prices per the vendors' pricing pages with annual billing, net of VAT.
When Zendesk is enough — the honest answer first
If your "portal" is at its core a ticketing system with a knowledge base and a status page, buy. The per-agent pricing model is fair for this case: your end customers don't cost a license, only your employees do. A quick example: eight service agents on Zendesk Suite Team cost 8 × €55 × 12 months = €5,280 per year. No custom development beats that price — and you're live in days instead of months, with a mature interface and an ecosystem of ready-made integrations.
Buy is also the right call when your service processes aren't standardized yet. Custom software cements processes — digitize chaos and you get digital chaos, just more expensive. And if you're not even sure yet whether you need a portal, a web app, or simply a better website, our decision guide Website, Web App, or Portal? settles that question before any make-or-buy discussion.
Where buy gets expensive: three cost drivers that aren't on the pricing page
ERP depth. Helpdesk SaaS shows tickets and help articles — not your orders, invoices, delivery dates, or stock levels from the ERP. Yet that's exactly what B2B customers ask about most. In the SaaS model, that connection becomes a custom project on someone else's platform: marketplace apps, middleware, consulting days — costs that appear on no pricing page. We've covered the available integration patterns and what your legacy system needs to support them (spoiler: not much) in our article on SSO and ERP integration for customer portals.
The add-on staircase. The entry price is not the price growing organizations pay: the jump from Suite Team to Suite Professional means €115 instead of €55, AI assistance (Copilot) adds €50 per agent, telephony via the Contact Center adds €82 — the path from the €55 entry point to a full build-out is a multiplication per agent, and the enterprise prices aren't even on the website.
Outcome pricing at volume. Billing per automatically resolved ticket sounds fair, but it permanently ties your costs to your request volume. Example: 2,000 automatically resolved tickets per month at 1.50 US dollars per resolution cost 3,000 US dollars monthly — year after year, on top of the licenses. Your own portal answers the same requests at a marginal cost of roughly zero.
The decision framework: five criteria
Instead of feature comparisons, we recommend five criteria you can work through with IT and the business side in a single workshop afternoon:
| Criterion | Points to buy (SaaS) | Points to make (custom portal) |
|---|---|---|
| Process standardization | Your service processes match the industry standard — and can stay that way | Processes are customer-specific or part of your differentiation |
| ERP/CRM integration depth | Tickets, FAQ, and knowledge base are enough; ERP data stays out of scope | Customers should see orders, invoices, delivery dates, and stock levels directly |
| User count and license trajectory | Few agents, manageable ticket volume; end customers don't cost a license | Growing agent count, add-on needs, high volume under outcome-based AI pricing |
| Data sovereignty and GDPR | US cloud is contractually and regulatorily acceptable | EU hosting, your own identity (SSO), and full data sovereignty are required |
| Differentiation potential | Support is a hygiene factor, not a competitive advantage | The portal is part of the product or a sales channel in its own right |
How to read it: if four or five checkmarks land in the left column, buy SaaS and invest the difference in your processes. If three or more land on the right, run the numbers on the make case. With a mixed picture, the hybrid approach below is usually the most economical answer.
The tipping point in numbers: a worked 5-year example
The following calculation is an example, not a study — and we lay out the assumptions: twelve service agents, list prices with annual billing (as of July 2026), no price increases over five years — a deliberately SaaS-friendly assumption. On the make side stand our published price anchors: €20,000–60,000 for a customer portal with roles and ERP integration, plus operations at €20–50 per month for infrastructure and two to five hours of support at roughly €120 each.
| Scenario (12 agents) | Cost over 5 years | Included / not included |
|---|---|---|
| Zendesk Suite Team (€55/agent/month) | €39,600 | Ticketing and knowledge base; no ERP data, no self-service at the order level |
| Zendesk Suite Professional + Copilot (€165/agent/month) | €118,800 | AI assistance (Copilot) included; ERP self-service still not |
| Freshdesk Pro + Freddy Copilot ($84/agent/month) | approx. $60,500 | Comparable helpdesk scope; additional AI sessions cost extra |
| Custom portal (make) | approx. €36,000–99,000 | Build €20,000–60,000 plus operations approx. €260–650/month; ERP integration, SSO, and EU hosting included |
Worked example, as of July 2026: list prices per the vendors' pricing pages with annual billing, excluding discounts and price increases; the make ranges are our published price anchors. All prices net of VAT.
The table shows two things. First: make is not automatically cheaper — over five years, Suite Team beats any custom development as long as its feature set is sufficient. Second: the comparison tips at two points. When add-ons and agent growth push the SaaS price toward a full build-out. And — more often — when ERP self-service is required: then you pay SaaS licenses plus an integration project on someone else's platform, while the custom portal already includes the ERP integration. Paying twice for half a solution is the most expensive outcome of this decision.
The third path: hybrid
In practice, a combination often wins: the SaaS helpdesk stays for tickets and the knowledge base — that's where it's good and cheap. Alongside it, a lean custom portal covers what SaaS can't: order status, invoices, documents, and reorders pulled straight from the ERP, connected via shared single sign-on. A portal MVP like that runs €8,000–20,000 with us and validates the business case against real user behavior before the big investment lands.
McKinsey shows this pays off as more than cost reduction: according to the B2B Pulse 2024, 39% of B2B buyers are willing to spend more than 500,000 US dollars per order through self-service and remote channels — up from 28% in 2022. And according to the Gartner Sales Survey (published March 2026), 67% of B2B buyers now prefer a rep-free buying experience. How to turn that into a solid internal business case is covered in our article on the business case for a customer portal.
How to decide now
Three steps you can take this week. First: calculate your SaaS costs over five years — agent count times list price, honestly including the add-ons you'll need in two years. Second: list your ten most frequent recurring customer requests and mark which ones need data from the ERP — that's your make indicator. Third: walk through the five criteria from the table together with IT and the business side — not sequentially, but in the same meeting.
If the answer turns out to be make or hybrid: we start every portal project with a make-or-buy check during discovery — and we'll still tell you if Zendesk is enough. What a custom-built customer portal costs and how we build it on Next.js, Supabase, and Keycloak is published openly on our service page. Or bring your own 5-year calculation straight to a free initial consultation — 30 minutes, no slides.
