Conversational access to business applications: 10 criteria for product managers
Thierry Grenot
In this article we share with you the criteria to consider when integrating conversational AI with business software.
The release of chatGPT to the general public in December 2022 highlighted the use of natural language. This is the language we use in everyday life. To carry out complex searches and obtain answers that are easy to use…
A simple perspective shows an inflection point for business software user interfaces. And this applies regardless of the application in question: HRIS, ERP, Finance, Payroll or Resource Management…
This tipping point is the third generation of interactions.
Web, mobile applications and conversational AI
• Encounter of the first kind: the Web
The development of technology in the 1990s (increasingly fast Internet, Web, JavaScript, XML…) enabled the advent of the Cloud at the very beginning of the 21st century, at the same time as the widespread use of Web interfaces for business software.
• Encounter of the second kind: Mobile applications
The arrival of the first iPhones, with their “store” of applications (2007), opened up the possibilities of rich, evolving mobile access. They made possible a revolutionary and rewarding experience for users.
• Encounter of the third kind: Conversational access
Exprimer sa demande au moyen d’une conversation simple et directe. Cela permet gain de temps et de confort pour l’utilisateur. Plus particulièrement, pour les usages occasionnels ou les métiers de première ligne, la conversation est l’accès le plus évident. Pas de formation, fonctionne depuis toute application de messagerie professionnelle ou personnelle, multilingue, vocal, etc.
More than just a fad, conversational interfaces offer long-term, high-impact benefits. And business software publishers and their customers alike are reaping the benefits:
- Quality of experience, the key to success: high functionality, competitive pricing and quality support are no longer enough to make the difference.
- Adoption rates to guarantee the low attrition necessary for the success of SaaS solutions.
By offering intuitive features and rapid access to solution functions, users become more autonomous. The direct benefit is to lighten the load on customer service.
Innovative positioning, easy to demonstrate and focused on the simplicity and ease of use of your application, strengthens your image in the market and increases the impact of your marketing and sales actions.
which criteria should be considered to choose a conversational AI?
Here we outline 10 criteria that teams responsible for defining and developing enterprise applications could take into account to ensure the success of their conversational AI projects:
- Use cases
- Flexibility
- Access channels
- Multilingual
- Time to market
- Cost (development, maintenance, licenses…)
- Available skills
- User experience
- Explainability
- Security, confidentiality, sovereignty
Want to know more? We’ll give you more details on these criteria for analyzing conversational AI solutions.
10 criteria to consider for conversational AI
1. Use Cases
Usage naturally depends on the functionalities of the enterprise software in question, as well as on the profession of the target populations. Identifying conversational usage is generally not very difficult. The starting point is the functions offered on the web portal and/or mobile application (or even the operational support team). And all functions that do not require a complex graphical interface are eligible.
For a complete product (HRIS, ERP, Finance…), we can generally identify a few dozen relevant uses for a conversational interface.
2. Flexibility
Les besoins clients évoluent sans cesse, et vos compétiteurs de même. La gestion de votre roadmap comporte des réglages fins, avec des approches progressives combinant aspects défensifs (ne pas perdre de clients, tenir les prix) et offensifs (en gagner de nouveaux, ajouter de la valeur).
Certains aspects piégeux ne doivent pas être oubliés : jargon métier, noms propres, authentification, etc.
Le langage naturel possède toutes les caractéristiques pour s’adapter aux futurs besoins de vos clients. Il reste à s’assurer que l’implémentation sera à même de les supporter sans effort et à moindre coût.
3. Access channels
Vos utilisateurs sont déjà sur les réseaux sociaux et sur leurs applications collaboratives (outre votre site web et application mobile). Il serait dommageable de ne pas profiter de cette diversité pour maximiser les accès à votre solution, et ce tout en offrant une expérience unifiée.
Selon le contexte, vous souhaiterez être présent sur MS Teams, Slack, WhatsApp, votre application mobile, etc… Voire en vocal.
4. Multilingue
Là c’est simple, vous savez déjà quels sont les langages de vos utilisateurs. Ce qui l’est peut-être moins, c’est la capacité d’offrir des accès conversationnels simultanément à tout votre parc client, et de pouvoir ajouter “à volonté” de nouvelles langues.
5. Time to market
Sauf si votre équipe possède déjà des compétences pointues en traitement automatique du langage (TAL), ne vous attendez pas à des résultats rapides.
C’est un sujet piégeux, y compris en présence d’outils apparemment magiques tels que les LLMs (ou large language model). Et comme toujours, c’est le passage d’une maquette spectaculaire à un ‘vrai’ produit qui sera le plus difficile.
Selon le périmètre que vous sélectionnerez et les partenaires que vous impliquerez (ou pas), il faudra le plus souvent compter entre 6 et 24 mois pour une généralisation.
6. Cost of a Conversational AI
- Design and development: directly linked to your choice of partners, technologies and in-house development. Probably between €60k and €600k, depending on the case.
- Operation: this point is often overlooked. But, unsurprisingly, NLP and generative AI “LLM” tools are not free. Inadequate implementation can lead to significant operating costs – beware of the long-term impact on financial equilibrium!
- Maintenance: not to be forgotten, of course.
7. Available skills
Development teams are neither omniscient nor always available. They have to deal with corrective maintenance, functional additions requested by customers – and those desired by sales and marketing teams, etc. Product and R&D managers then have to arbitrate in the midst of contradictory injunctions.
Product and R&D managers must arbitrate in the midst of contradictory injunctions. They seek to maximize the impact of their actions for the benefit of the company. In most cases, this means concentrating forces on the core business (the application) and the quality of the user experience, while finding ingenious solutions for innovation.
8. User Experience
Familiar with using chatGPT, users have a high level of expectation. This is sometimes even higher than what would be expected from a human interlocutor… Some criteria to consider for a satisfactory UX :
- Quality of understanding
- Tolerance of typographical errors
- Quality of responses
- Reaction to erroneous requests
- Help for the user
- Absence of hallucination
- Speed of exchanges
9. Explicability of AI
NLP, no matter how much AI goes into it, is not magic. But it can be opaque. Particularly when training data is unknown. Just like a human interlocutor, a NLP can have biases of all kinds. It can be more or less ill-intentioned, or even fall victim to hallucinations or “fake news”.
Mastery of the various technologies and of the data used can provide the required transparency.
10. Security, privacy, sovereignty (or the main criteria for choosing a conversational AI)
Enterprise applications often handle confidential or even personal data. To meet customer expectations, you need to know how to handle this information, while complying with legal obligations and security rules.
Anonymization, hosting location and legislation, database management, user and administrator rights… everything counts. A “security & privacy by design” technical architecture will enable you to achieve the required level of protection.
The evolution of technologies and available solutions means that conversational interfaces are already accessible to all business software publishers. But this functionality requires skill, know-how and attention.
After all, it’s your future point of contact with your customers.
Did you like this article?
Agora Software offers conversational application interfaces, particularly adapted to business software. We create rich, multilingual and omnichannel interactions with all your users.
Do you have a conversational interface project?
Let’s talk: contact@agora.software
Want to understand how our conversational AI platform optimizes your users’ productivity and engagement by effectively complementing your enterprise applications?