1/ AI has become a new paradigm in the Software Publishers ecosystem

Seizing the opportunities offered by AI is no longer just a marketing requirement. It’s now a matter of survival for software publishers. While the SaaS market has increased and has remained buoyant ever since COVID pandemic, we have seen companies streamlining the number of SaaS services they use. They had too many services (269 applications on average per company) that are still largely underused.^[1]

In this context, AI is seen as a way to make a difference. How? By consolidating the added value of software and enhancing its features. This is all the truer given that you no longer need to be a so-called “DeepTech” company to use AI. Now, any software publisher can now have a go at it, without any specific skills. In many cases, it is possible to leverage pre-trained AI algorithms to infer the model with its own data and produce results. The Franco-American startup Hugging Face, already known as the “Github of AI”, has made available more than 500,000 open-source artificial intelligence models and 250,000 sets of data.

Are you ready? Here’s what AI can bring to your application development.

2/ Organising and enriching data to better exploit it

Thanks to the combined advances of machine learning, natural language processing (NLP), computer vision and the advent of vector^[2] databases, one of the new capabilities of AI is to organise your data so that it can be immediately used: sorting, indexing and even enriching it. What previously required the rare skills of data scientist profiles is now easily within reach.

Carried out from the collection phase or retrospectively on older data, data pre-processing by AI opens up possibilities for immediate utilisation of your data, starting with AI-powered search. AI’s ability to understand both the user’s request (prompts written in natural language), and indexed content is a significant contribution to research. The aim here is to display the expected answer by directly synthesising the information and attributing it to its source, instead of using ranking criteria. This revolution was led by a long-time OVHcloud customer, the French platform Algolia. Is the next challenge for OpenAI to compete with Google in search engines?^[3]

The Retrieval-Augmented Generation (RAG) technique optimises the results of large language models (LLM) by connecting them to specific resources, without retraining the model. In short, it’s about training an AI model with a well-defined corpus of external and/or internal data that can be added on the fly, to avoid off-topic or unsourced results and other AI hallucinations.^[4]

There is, however, a barrier to model training and specialisation via RAG: the lack or difficulty of exploiting certain data due to regulatory reasons. Here, too, AI could prove to be a game-changer, with the creation of “synthetic data”^[5]: artificial data with the same characteristics as the original data by sampling and modelling their probability distribution. It has the potential to transform the way we consume, exchange and monetise data.

3/ Making sense in real time and predicting/reducing uncertainty

AI significantly improves predictive modelling by analysing historical patterns faster and more efficiently. Predictions can be refined by basing it on a higher quantity of data. Where predictive scenarios can be multiplied by playing on an unprecedented number of factors. It also becomes possible to generate and adjust these predictions almost in real time. Idenifying trends and providing valuable insights to make more informed decisions faster and with less effort.

In the field of LegalTech, the French search engine Predictice, an OVHcloud client since its beginnings, is collecting case law from courts in opendata to feed algorithms that can provide predictions on the chances of a trial succeeding based on a multitude of parameters, with the aim of guiding lawyers in developing their defence strategies.

The marketing industry is also using AI to bring brands closer to influencers. By calculating the ‘affinity’ between the audience of an account and the service or product that a company needs to market. This will then help ‘micro-influencing’ or even ‘nano-influencing’ emerging.

Applications for predictive AI are virtually limitless. Whether it’s detecting fraud, optimising clinical or consumer testing, anticipating sales changes to optimise a production line and tie up less capital in inventory, organising predictive infrastructure maintenance based on data collected by connected sensors. The list goes on.

4/ Revolutionising the man-machine interface and supporting developers

Beyond the obvious applications, for example in customer service, chatbots have become a new way of accessing data and querying it using natural language. “Prompts” enable the streamlining of application interfaces, or generation of dashboards on the fly, which eliminates the need for coding or clicking.

In short, AI could renew interest in “progressive disclosure”, a user-interface design concept popularised in 2000 by Jef Raskin in his book The Humane Interface,^[6] which is based on three principles:

• Initial simplicity: at first glance, only critical features or information are visible

• Easy access to additional information as needed by the user

• Customisation: users can adapt the interface so that it better responds to their needs and preferences (in this case, AI has the ability to do this even without the user’s intervention).

AI is changing the landscape of software development. From simple wizards to tools like GitHub Copilot, developers now have resources to create, explain, enhance, correct, and optimise their code. Both beginners and veteran developers can use these tools.

While “no code” is still a pipe dream for programming complex applications, “low code” has become the norm. Shouldn’t a developer focus more on assembling tech building blocks rather than mastering a programming language?

The combination of low Code and DevOps is particularly effective for quickly bringing an idea to fruition and boosting innovation capacity. Especially since AI also automates some of the testing, and speeds up the deployment of software.

5/ Hyper-personalisation of the customer/user experience

We did not wait for AI to start building recommendation engines. Indeed, the fields of e-commerce, music or video streaming have known a huge success. You are probably wondering what’s new is the power of customisation that AI can enable? Well, you will be surprised. It’s no longer about building groups of users with similar behaviours. It has now become possible to personalise recommendations for a single user, leaving behind the era of mass marketing.

Best of all, this customisation, coupled with generative AI, could lead to a generation of unique, individualised content. A brand might consider creating a marketing campaign whose visuals are generated from the company’s advertising history (previous campaigns, to teach the AI model the brand’s communication codes) and adapted to the user’s tastes.

And the gaming world is no stranger to this, as it is already thinking about using AI to boost interaction from non-gamers. Even astrology, which up until now has been sufficiently vague so that everyone can find a grain of truth within it, cannot resist the siren song of AI-assisted hyper-personalisation.^[7] This hyper-personalisation also brings promises in the fields of education, training or even health. Each one shares a future where one of its cornerstones will be individualised outputs, such as educational content, treatments or health advice.

6/ Assist, optimise, automate: from “Micro SaaS” to creating AI agents (agentisation)

The last big family of AI breakthroughs is probably the least spectacular in appearance, but it is undoubtedly the one that is currently driving the most innovation among SaaS publishers. In this new generation of applications, AI is permeating everywhere and playing a central role, not only as a support tool, but also as a robot capable of making decisions and executing tasks without human intervention.

In this day and age, what could be more surprising than the number of hours companies waste on Excel or PowerPoint to perform tasks that AI algorithms can now perform effortlessly?

AI has already had the capacity to handle multiple time-consuming and repetitive operations for a few years. From analysing data to generating reports, optimally planning a tour or writing a draft response to an email, AI can generate significant productivity gains within enterprise applications.

To serve its legal sector clients, Septeo Group has made AI a top priority, offering them end-to-end solutions for drawing up contracts and legal documents. The human resources industry also provides some examples of software that boosts their users’ productivity by helping them with everyday tasks.^[8]Tortus.ai, a chatbot capable of assisting a doctor during a consultation by summarising it for the patient’s record and suggesting the contents of prescriptions for further examination, shows how far automation can go, even in a field known to be complex.^[9]And for a more general purpose, the Automatic Document Reading (ADR) solution developed by Recital.ai gives an idea of future developments in electronic document management.

Finally, the trend toward “micro-SaaS”, referring to lightweight (or even cobbled-together) software solutions that serve highly targeted needs, often with the help of AI-based’ tools, shows that there are still many gaps in the professional applications market. There is ample room for improvement, and it doesn’t have to cost a fortune.

Alongside this trend of micro-SaaS, the “agentisation” of AI services appears to be an ideal solution for automating complex tasks, by using an agent to organise a workflow that chains the processing of several services through their APIs. Enough to work miracles – or almost, while we wait for a hypothetical super-intelligent entity capable of organising all these based on a simple scheme of intent.

Careful, quick time-to-market doesn’t necessarily mean acceleration

In a rush to move swiftly, there is a tendency to choose the fastest option, which involves relying on pre-built AI services and endpoints – offered directly by companies that develop widely used general-purpose or specialised AI models. Watch out for potential risks! Such as data leaks and unauthorised use of your valuable data to train models that don’t belong to you. It could ultimately benefit your rivals. This also includes the risks of getting locked into non-reversible services. Our next blog post will explore the risks involved, the state of European regulation, and the advancement of “trusted AI”.

In the meantime, you can rely on OVHcloud’s services to safely test and innovate with AI. OVHcloud, the leading European cloud provider offers various solutions for training, deploying and inferring models based on your expertise. Our offerings include on-demand GPU servers, pre-trained models to quickly integrate features into:

– your applications (AI Endpoints),

– ready-to-use work environments (Notebooks as a Service) for your developers and data scientists,

– AI Training and AI Deploy services for orchestrating compute resources to suit your needs.

Find out how to use OVHcloud AI Machine Learning to fuel your business: https://www.ovhcloud.com/en-gb/public-cloud/ai-machine-learning/

^[1] https://www.cio-online.com/actualites/lire-le-saas-une-poche-de-depenses-a-optimiser-pour-la-dsi-15510.html

^[2] https://blog.ippon.fr/2023/08/30/les-bases-de-donnees-vectorielles-2/

^[3] https://www.usinenouvelle.com/article/openai-va-devoiler-un-engine-de-recherche-to-compete-google.N2212813

^[4] https://en.wikipedia.org/wiki/Hallucination_(artificial_intelligence)

^[5] “Sky Engine AI raises $7 million for its synthetic data generation cloud solution”:

https://www.usine-digitale.fr/article/sky-engine-ai-leve-7-millions-de-dollars-pour-sa-solution-cloud-de-generation-de-donnees-synthetiques.N2206767

^[6] https://lagrandeourse.design/blog/actualites/progressive-disclosure-simplifier-linterface-pour-mieux-guider-lutilisateur/

^[7] https://madame.lefigaro.fr/astro/l-intelligence-artificielle-est-elle-le-futur-de-l-astrologie-20231231

^[8] https://meetcody.ai/fr/blog/7-principaux-outils-dia-saas-pour-les-rh-en-2023/

^[9] To discover more applications of generative AI in healthcare, read this article: https://www.mind.eu.com/health/essentiels/ia-generative-ou-sont-les-opportunites-en-sante/

Since the dawn of cloud, customers have spent hours, days and weeks deciding which cloud environment is right for them and their IT estate. The channel has been an intrinsic part of this evaluation, with partners supporting their customers and helping them make the right decision for their own specific needs.  

Today, cloud is not only a mature technology, but it’s an interoperable one as well. Organisations use multiple clouds for different applications and workloads, selecting best of breed vendors to fit their needs. This multi-cloud approach is fast becoming the norm. We recently surveyed over 500 technology buyers from the UK, and almost two thirds (62%) are actively using multi-cloud environments today. A further 23% are either transitioning to multi-cloud or exploring the possibility of multi-cloud in future.  

Although there are a number of organisations that do use a single cloud provider, a staggering 65% of organisations say that their use of multi-cloud will increase in the next two years, representing a huge opportunity for the channel. Mixing and matching workloads to the right cloud environments can bring benefits, but also brings more complexity, with integration into other applications and a greater number of endpoints to secure. Around a third (31%) of IT decision makers said that they were concerned about securing their multi-cloud environments, and 27% understood that multi-cloud would make their estates considerably more complex. 

So what does this mean for the channel? How do channel partners adapt to address these market dynamics? Partners can provide invaluable experience and value to customers, helping to tame their multi-cloud environments. By getting to know a customer’s infrastructure, helping them to choose the right vendors and giving good consultancy, channel partners can not only help customers transition to multi-cloud effectively, but build both long-term relationships and long-term margin, in particular by adding complementary support and consultancy services.  

We’re at a key inflection point and the time to act is now. Three quarters of respondents stated they need more support and expertise, with only a quarter of organisations see multi-cloud as ‘plain sailing’. Around half (46%) are ‘getting there, one step at a time’, with a fifth (20%) admitting that if multi-cloud was a relationship status, theirs would be ‘it’s complicated’.  

But whether your customer is a mature multi-cloud user or just starting down the road, it’s almost certainly a journey that they won’t diverge from, and as their complexity grows, so will a desire to seek help from channel partners. The benefits of multi-cloud are well understood: in our poll of over five hundred customers, only one believed that there were no benefits to multi-cloud. Everyone else knew that using the right cloud for the job could give you better flexibility, better portability, lower risk and better cost-effectiveness.  

And with those kinds of benefits on offer, it’s quite clear that we’re going to be living in a multi-cloud future, and the channel opportunities are there for the taking.