Conversational AI
Virtual assistants and chatbots powered by artificial intelligence
We build conversational assistants that interpret the user’s query and resolve it from the organisation’s corporate documentation, citing the source document in each answer and referring the query to the relevant team where professional judgement is required.
Areas of application
The assistant resolves queries from the organisation’s manuals, internal policies, catalogues or incident history, and indicates the source of each answer.
Immediate resolution of the highest-frequency queries, available on a continuous basis, with referral to the support team in the cases that require it.
Direct consultation of procedures, internal policies and operational data, without having to locate the information in shared repositories.
In addition to resolving queries, the assistant collects the required data, verifies that it is complete and generates the ticket, booking or request in the corresponding system.
Web, mobile app, WhatsApp, email or internal messaging tools, with consistent behaviour across all enabled channels.
Each answer is accompanied by a reference to the source document, allowing the information provided to the user to be verified at any time.
Substantiated, verifiable answers
Our assistants answer from the organisation’s corporate documentation and accompany each answer with a reference to the source document, available for immediate consultation. The scope of what the assistant resolves autonomously is agreed with the organisation in advance, and queries beyond that scope are referred to the relevant team with the case details already collected.
Reference to the source document
Each answer includes a link to the document and section the information comes from, accessible to the user at the point of the query.
Agreed scope and referral
We jointly define the categories of query the assistant resolves and those passed to a member of staff, together with the referral route.
Logging and continuous improvement
Conversations are logged and reviewed periodically to extend the assistant’s coverage and refine its answers.
Privacy and control of information
The treatment of internal documentation is defined prior to development: which information is incorporated into the system, where it is processed and which users may consult each type of content, respecting the permission scheme in force within the organisation. All developments comply with the GDPR and rest on our information security management system certified to ISO/IEC 27001.
Permissions inherited from source systems
The assistant respects each user’s permission scheme, so that the information accessible through it matches what is available through ordinary channels.
Minimisation of personal data
Only the information necessary for the use case is incorporated into the system, with anonymisation or pseudonymisation prior to processing where applicable.
Deployment model
Depending on the nature of the information, the assistant may be deployed on the client’s infrastructure or in environments where data is not used to train third-party models.
Working methodology
Analysis of recorded queries
We review the history of tickets, emails or calls to establish the type and frequency of queries, the basis on which the project is sized.
Preparation of the document base
We organise and structure the documentation the assistant will operate on, and assist in identifying the versions in force.
Development and testing on real cases
We test the assistant’s behaviour against historical queries and the cases identified by the organisation, until performance is consistent.
Progressive rollout and tuning
Deployment begins by query category or user group and is extended as the review of logged conversations progresses.
FAQs
Frequently asked questions about virtual assistants
How does this differ from a traditional chatbot?
An assistant built on language models interprets the user’s intent irrespective of the phrasing used and operates on the organisation’s corporate documentation, without each answer needing to be programmed in advance. The resulting coverage is substantially greater than that of a chatbot based on rules and keywords.
How is the accuracy of answers ensured?
The assistant answers exclusively from corporate documentation and includes a reference to the source, allowing each answer to be verified. Categories of query requiring professional judgement are referred to the relevant team along the agreed route.
What documentation is needed?
The organisation’s documentation in force: manuals, procedures, internal policies, catalogues or query history. In the initial phase we assist in organising it and identifying the applicable versions.
Does it replace the customer service team?
It redistributes their workload. The assistant handles the highest-frequency, lower-complexity queries, allowing the team to concentrate on cases that require professional judgement and have the greatest bearing on customer satisfaction.
How soon can it be operational?
An assistant with a defined scope over organised documentation enters service within a few weeks, with progressive rollout by query category or user group.
Which languages can the assistant operate in?
Whichever the organisation requires. Language models operate natively in multiple languages, so a single assistant handles queries in Spanish, Catalan, English or others from the same document base, and answers in the language it is asked in.
ALS Innovation
Tell us which categories of query account for the greatest volume in your organisation and what documentation is available. We will analyse the most appropriate scope for the assistant.