For the vast majority of companies, buying a SaaS solution, meaning an AI chatbot sold on subscription, is cheaper and faster than building one in-house. According to our 4-year simulation, for 100,000 conversations a year, an in-house project costs €444,000, compared with €105,000 for a SaaS solution. Here is what you need to decide whether to build or buy a chatbot.

Since ChatGPT arrived, one question keeps coming up: why not build our own conversational agent, connected to our tools? Yet an effective AI chatbot is not a side project: it is a complex product to design, maintain, secure and monitor.

Key takeaways

• An in-house project ties up a full team, and its costs are often underestimated.

• An in-house chatbot goes live in 8 to 12 months, a SaaS solution in 4 to 8 weeks.

• Fewer than 2% of French companies meet the four conditions that justify in-house development.

An in-house project ties up resources at every stage

Even in an optimistic scenario, an in-house chatbot project ties up human, material and organisational resources at each of its six stages. These costs are often underestimated. Delays and unplanned changes easily inflate the budget and push back the return on investment.

• Building the team: product managers, engineers, IT managers, security, compliance.

• Scoping workshops with the business teams and the IT department: scope, requirements, use cases. After a dozen or so meetings, the team finally has its specifications.

• Initial technical tests, to validate feasibility and answer quality.

• Development: coding, testing, adjustments, internal sign-offs.

• Go-live, often delayed, because the project is not a priority for the teams involved.

• After launch: user feedback, fixes, constant adjustments.

Building or buying a chatbot: what are the practical differences?

Large language models (LLMs), the AI systems that can understand and write text, now make it possible to automate some human tasks. Yet many teams underestimate the resources needed to reach the level of the best solutions on the market. The table compares the two options.

In-house chatbot or SaaS solution

*Performance measured across ViaSay customers over the last 12 months.

The autonomous resolution rate is the share of requests the chatbot resolves without an agent. Improving it by 20 to 30% can be worth the equivalent of 1 to 2 agent positions. At €45,000 per agent per year (salary, social charges, training, tools), the gap quickly adds up to tens of thousands of euros a year.

Over 4 years, an in-house chatbot costs €339,000 more

Take an in-house project with no major setbacks, for a customer service team handling 100,000 conversations a year. Over 4 years, it costs €444,000, compared with €105,000 for a SaaS solution such as ViaSay. Here is the breakdown, to rework with your own figures.

Development phase (6 months)

• Team: 1 project manager, 1 product manager, 2 senior engineers.

• Fully loaded cost: €75,000 per person per year.

• Development cost: €150,000.

Recurring costs (per year)

• Infrastructure and language model APIs: €36,000.

• Maintenance and updates (0.5 full-time equivalent): €37,500.

Budget comparison over 4 years for 100,000 conversations a year

👉 Total difference over 4 years: -€339,000

Four conditions justify building your chatbot in-house

In-house development is only justified in rare cases, when four conditions are met. Fewer than 2% of French companies meet them. Even well-funded tech start-ups regularly fail, for lack of AI expertise or because they underestimate the operational complexity.

1. Top-level, native technical expertise. The company has had experienced AI and machine learning (ML) teams for several years.

PayFit, the French HR unicorn, fits this profile: 841 employees, 124 of them engineers. The company recently built its AI assistant successfully, but it started simple: questions and answers on a knowledge base, with no complex connection to its systems.

2. Critical volume and a substantial budget. At least 500,000 interactions a month and a budget of more than €500,000 a year. Below that, a specialist team is not cost-effective compared with SaaS solutions.

3. Binding regulatory constraints. In some highly regulated sectors, data legally cannot leave the internal infrastructure. Even then, solutions installed on-site (on-premise) often exist.

4. Strategic differentiation. The chatbot must be a direct competitive advantage, not just a lever for profitability and efficiency. This is the case for a software vendor whose offering is built on conversational AI.

In most cases, buying protects both the customer experience and the budget

In the vast majority of cases, building your own AI chatbot can degrade the user experience and delay the return on investment. Costs can reach 3 to 4 times the initial estimates. A proven solution lets your teams focus on their core business, delivers a quality experience from day one and makes the budget predictable.

Considering a ready-to-use solution?

ViaSay offers AI chatbots that are easy to integrate, customisable and quickly up and running. Our customers achieve an average autonomous resolution rate of 75% in less than a month.

👉 Request a personalised demo today.