Read Time: 3 minutes
Generative AI is gradually arriving in various forms in the enterprise. In particular, it is being integrated into packaged software applications such as GitHub Copilot.
Through our numerous GenAI project implementations in 2023, we are convinced that one of the major challenges for businesses in 2024 will be the adoption and integration of a Generative AI platform.
These platforms centralise multiple use cases, facilitate cross-team collaboration on various projects, provide RAG-type processing chains and connection to AI models, whether classical or generative.
They also play a crucial role in testing and optimising the performance of AI models.
Definition
There are several types of platform in companies: data & AI platform, conversational platform, process automation platform, etc.
A Generative AI platform brings together a number of the characteristics of the above platforms to create, configure, evaluate and deploy software solutions augmented by generative AI models. The software solutions may take the form of an AI Copilot, Chatbot, intelligent search engine or augmented advisor.
In addition, these platforms, which incorporate access to generative AI models, also make it possible to use 'classic' AI models to perform complementary operations.
Top trends driving the emergence of platforms
The emergence of platforms dedicated to generative AI is being driven by several key trends in the current technological landscape:
1) The number of generative AI models is multiplying.
There are no winners in the so-called foundation models, and open-source models are gaining in popularity.
2) AI is no longer the preserve of data scientists.
The growing demand for generative AI from business users is revealing a growing diversity of use cases, ranging from business process optimisation to product and service innovation. Companies want to move quickly from MVP to industrialisation.
3) Variety of data
These business scenarios often require the use of a variety of data within a heterogeneous application ecosystem, underlining the importance of a platform capable of orchestrating and integrating.
3) Data accessibility
The effectiveness of generative AI is highly dependent on access to and use of enterprise data, whether stored in transactional systems such as Salesforce, SAP, ServiceNow, or in knowledge bases such as SharePoint, Confluence, or websites.
4) The need for testing tools
Finally, to maximise the effectiveness and relevance of the results produced by AI, companies need to have the right tools to test and optimise AI processing chains and minimise model errors.
The challenges facing AI gen platforms
Be agnostic with regard to LMM suppliers
The number of LLMs has been rising sharply for several months and a battle is raging between Open Source models (Mistral.ai, Llama2, etc.) and closed source models. The GenAI Studio architecture needs to be agnostic with regard to the various suppliers of generative AI models, enabling effective abstraction of the underlying models.
Keeping costs under control
Managing and optimising the costs associated with the use of generative AI models is a major economic challenge, directly affecting the return on investment. Indeed, if a project player calls upon an AI model on large quantities of data, and on numerous occasions, this can be very costly for the company.
Integration, security and observability
Seamless integration of AI solutions into existing systems, coupled with robust security measures and full observability, is crucial to ensure seamless operation.
Performance testing, monitoring and optimisation
The implementation of rigorous protocols for testing and monitoring AI systems, as well as access to advanced tools for improving performance, are essential for guaranteeing the reliability and effectiveness of models.
Agility and scalability
The ability to adapt quickly to change and to scale up to an industrial level is essential (version management, devOps processes, auditability, etc.).
Collaboration
Facilitate collaboration between teams from different disciplines, working on several generative AI projects. This requires granular management of the roles and permissions of those involved in the project.
Host
Depending on the case, the platform must be available onPrem, in a private cloud or in the public cloud.
The generative AI platform at the heart of the company's information system
Components
An effective generative AI platform is made up of several key components.
Workflow and orchestration engine
This component is essential for configuring and orchestrating business processes and data processing chains. Depending on the platform, ready-to-use workflows and processing chains may be available.
AI engines
The platform must integrate :- Classic AI engines (clustering, classification, translation, etc.),
- LLM models (text, image, code)
Search/retrieval function
Necessary to augment LLM models with your enterprise data.
Test and optimisation tools
These tools are used to facilitate the evaluation and debugging of machine learning models.
Ready-to-use user interface
Intuitive user interfaces (Copilot, Chatbot, search engine, augmented advisor) that are easy to customise.
API
An open platform for seamless integration and two-way data exchange between the GenAI platform and other systems or applications.
Dashboards
Dashboards to analyse the performance of the platform.
SDK
Required to extend the platform via code, for example to integrate additional libraries (e.g. LangChain).
Conclusion
The emergence of generative AI platforms on the market represents a concrete response to the problems faced by businesses, which are faced with growing demand for more intelligent business solutions, in a more complex environment due to the proliferation of AI models, the security of data exchanged with models, ecological impacts and changes in regulations.
Companies now need to define their strategy and carefully weigh up the pros and cons of building or buying. They can count on an ecosystem of French software publishers that embraced the generative AI revolution early on. Konverso is part of this French ecosystem and offers a generative AI platform, certified SOC2 Type2, available in the cloud/on prem, agnostic model, with ready-to-use solutions and APIs for rapid connection to the customer's IS. It is now possible to have a complete French stack including hosting in France, a French generative AI model and a French generative AI platform.
To book a demo
https://www.odigo.com/wp-content/uploads/2021/10/Aberdeen-RR-Agent-Desktop.pdf https://www.zendesk.com/in/blog/ai-for-employee-experience/ https://www.gartner.com/en/newsroom/press-releases/2023-05-03-gartner-poll-finds-45-percent-of-executives-say-chatgpt-has-prompted-an-increase-in-ai-investment https://emplifi.io/resources/blog/customer-experience-statistics https://www.pwc.com/us/en/services/consulting/library/consumer-intelligence-series/future-of-customer-experience.html https://mytasker.com/blog/unexpected-collaboration-of-artificial-human-intelligence#:~:text=The%20Need%20For%20Collaborative%20Intelligence&text=As%20per%20Harvard%20Business%20Review's,most%20significant%20improvement%20in%20performance.&text=What%20comes%20naturally%20to%20us,a%20joke%2C%20and%20be%20sarcastic. https://www.bcg.com/publications/2023/how-generative-ai-transforms-customer-service https://www.forrester.com/press-newsroom/forrester-cx-na-event-agenda/#:~:text=Forrester's%20data%20shows%20that%20customers,more%20when%20companies%20communicate%20clearly. https://media.bain.com/Images/BAIN_BRIEF_Five_disciplines_of_customer_experience_leaders.pdf https://www.accenture.com/us-en/insights/song/accenture-life-trends
