The questions that you need to ask when selecting a Chatbot Vendor

Previously many Chatbot implementations were a disappointment to customers, primarily because they were either deployed for inappropriate use cases, were developed by IT and did not represent what business departments had intended, the Artificial Intelligence and Natural Language Processing  were not up to scratch or the scenarios and intents were not properly captured.  In summary, there are systemic, strategic, commercial and customer issues that need to be investigated as this White Paper outlines.

Pricing model

Some Vendors charge per interaction.  This could seem appealing at first if a customer is unsure how many customers would use the Chatbot, but it is difficult from a budgeting point of view and over time the cost could become expensive and unpredictable.

Alternatively some Chatbots are advertised as open source and ‘free’.  Of course customers tend to get what they pay for and the support may be poor whilst the platforms themselves will generally have less functionality, be less user friendly and often aimed as tools for IT Developers.

The issue for this is that the customer may only see what the Chatbook looks like after weeks of development and it may not be how they envisaged or work in a live customer environment.  All this development work is expensive and takes away IT staff from other work they could be doing.  Moreover, any changes to the system will require more coding, expense and delays.

Some may also be offered as an option for a specific tied platform (e.g. Hubspot CRM).  However, whilst this might be OK if you are just capturing Sales enquiries if you have other use case requirements then customers will have to work with more than one Chatbot supplier.

If a customer purchases a perpetual licence they will have the risk of a large upfront fee and increasing annual support fees.  Software-as-a-Service models enable customers to spread the cost of software and support into a monthly payment instead, reducing risk and maximising cashflow whilst offering relatively short contractual tie-ins.


Whatever use case you have in mind, integrating to a backend knowledge repository database will be necessary so the ease of integration will be key. Modern platforms will offer modern WYSIWYG capabilities for field mapping and support modern means of integrating such as Webhooks. Older technologies like database polling inevitably slow down performance significantly particularly for near real time technologies such as Chatbots.  The time it takes for doing implementations can vary substantially and be a significant element of the Total Cost of Implementation (TCO).


The level of support that you get may be limited for free or low-cost Vendors or be outsourced to offshore teams where there may be language issues.

As your Chatbot represents perhaps the first point of interaction for your customers it is essential that not only does the Chatbot work in an intuitive way but that any technical issues are resolved as quickly as possible.  If they are not then customer experience will suffer – the very metric that many companies are trying to improving with a Chatbot implementation!

As many customers stakeholders are involved in determining the use cases, intents and business drivers and are new to Chatbots then having a Chatbot Partner that can provide professional services to guide and structure this process ensures that the Chatbot is optimised.

With regards to Technical Support you would need to ensure that there are clear Service Level Agreements in place for uptime and response rates to trouble tickets because having a non-functioning or glitchy Chatbot will quickly annoy customers.

Just as importantly, the user interface should be intuitive for business users with minimal training, accompanied with necessary documentation, FAQs, training videos etc to enable customers to be confident in using the platform.

IT architecture

A Chatbot is only as good as the capabilities of the Artificial Intelligence Algorithms and Natural Language Processing engine that is being deployed.  As a customer you would need to know how the AI works and (if relevant) what foreign languages are being supported.  All AI platforms need to be trained but ideally the AI should be self-learning so that the more data it ingests, the more accurate are a Chatbot’s responses.  If the technology used is ‘black box’ then you would need to be able to stress test it yourself, ideally as part of a pilot implementation

Customers also need to look for Vendors that have a scalable and secure architecture which for SaaS Vendors will be tied to the hosting environment that they use e.g Amazon Web Services, Microsoft Azure, Google Cloud, IBM etc.  One of the advantages of these big Cloud hosting companies is that their architecture is inherently scalable, reliable and fast.

It is important to know where the data physically resides for some customer to comply with regulations like GDPR and whether there are adequate disaster recovery systems in place.

The Vendor should have analytics for metrics such as the effectiveness of the Chatbot, the number of successfully resolved queries, number of queries forwarded to agents so that the Bot can be optimised over time

Vendor Reputation

Before selecting a Vendor you should ask to speak to one or two customer references to learn about their experience.  Ideally the Vendor should be able to provide some references from IT Analysts but even a simple Google search enables customers to see reviews and employees’ perception of the company (Glassdoor).  Customers will be working closely with their Chatbot Vendor so it is important that there is a commitment from the Vendor to delivering excellent Customer Experience.