Low-code and no-code platforms
In our annual outlook, we predicted that low-code and no-code platforms in the field of conversational AI would receive a significant boost by 2025 – particularly as teams without in-depth development resources would also be able to create their own assistants. Looking back, it is clear that this development has made noticeable progress. Many platforms now offer visual editors, modular components and integrated LLM functions, meaning that simple chatbots or initial prototypes can be implemented much more quickly.
However, it has also become apparent that
In larger organisations or bot ecosystems that have evolved over time, purely low-code approaches continue to reach their limits. Complex integrations, editorially vetted knowledge sources, multilingual content, multi-channel delivery or sophisticated governance still require professional architecture and technical expertise. This is precisely where Materna’s hybrid AI assistant approach comes into play: an architecture that combines rule- and flow-based bot components with curated, structured knowledge and the natural language capabilities of modern LLMs. This enables existing systems and content to be reused in a controlled manner whilst simultaneously being enhanced with generative AI capabilities – without companies losing control over knowledge, integrations or quality. We have outlined the conceptual foundations and technical implementation in detail in our technical articles (‘Hybrid Chatbots – The Future of Customer Interaction’ and ‘Hybrid Chatbots – Architecture and Technical Implementation’).
Context
Low-code lowers the barrier to entry and accelerates the development of simple or clearly defined use cases. Hybrid architectures, on the other hand, combine the strengths of generative AI with the requirements for control, scalability and integration into enterprise landscapes. 2025 has thus demonstrated that the future of high-performance conversational AI solutions lies in the combination of both approaches – low-threshold creation and professional, robustly built AI systems.
Voice search optimisation
In our annual outlook, we had anticipated that voice search would continue to gain in importance in 2025 and that optimisation for spoken search queries would therefore also become more important. Looking back, the picture is more nuanced: voice search is present – but far less widespread than some forecasts from previous years had suggested.
What was actually achieved by 2025
According to the latest statistics, around 20.5 per cent of internet users worldwide now use voice search, and there are over 8.4 billion voice-controlled devices. Voice-based queries play a role primarily where quick, everyday information is sought – for example, in local ‘near me’ searches, in smart home environments or on the move via smartphone assistants. This usage is stable and has become even more established in 2025.
What has not yet become an established standard
Despite the prevalence of voice-enabled devices, many businesses are still not consistently optimising their content for voice search. Voice search has so far remained more of a supplementary channel, generating added value primarily in specific use cases. Traditional SEO strategies continue to dominate, whilst voice optimisation is used primarily for local listings, immediate service enquiries and hands-free situations.
Conclusion
This confirms that voice search optimisation is a relevant but selective trend. Companies that take natural language inputs, local search intent and structured data into account at an early stage will benefit – yet a comprehensive paradigm shift towards ‘voice-first’ will not have taken place by 2025.
Integration with IoT & Smart Devices
In our annual outlook, we had anticipated that conversational AI would become more prevalent in smart home and IoT environments by 2025, and that interaction with devices would become more natural. Looking back, it is clear that this development has gained significant momentum, but is still in a transitional phase. Many foundations have been laid, but the actual impact is likely to become apparent only in the coming year.
What was actually achieved in 2025
The major platform providers set a decisive course in 2025. Google began replacing its Assistant with the Gemini-based AI assistant, Apple introduced the first features of Apple Intelligence, and Amazon released Alexa+, a modernised version of its voice assistant. Manufacturers such as Samsung are also increasingly integrating AI-powered automations into their smart home ecosystems. Voice control is now established in many households and remains the primary form of interaction between users and connected devices.
However, this does not yet represent the breakthrough that was expected
Despite these advances, the limitations of the current generation are clearly evident. Many new features will still be rolled out gradually in 2025, are regionally restricted, or are only available on selected devices. Genuine generative AI capabilities – such as multimodal interaction, context-sensitive assistance or automated smart home workflows – have been announced, but are not yet widely available. Added to this is the continuing significant fragmentation of the smart home market, where differing standards, ecosystems and device categories make widespread and uniform use difficult.
Context
As such, 2025 was primarily a year of preparation: a multitude of announcements, new model generations and initial steps towards integration clearly point the way forward, but the actual shift towards truly intelligent, conversational smart home experiences will not become a reality until 2026. The trend is real and strategically important – but has not yet fully permeated everyday life.
Real-time multilingual conversations
Our forecast at the start of the year was that, by 2025, real-time translation would have advanced significantly and language barriers in digital interactions would increasingly be breaking down. Looking back, it is clear that the technology is evolving and has made noticeable progress – in both software and hardware – but a widespread breakthrough has yet to materialise.
What has actually been achieved by 2025
Several AI providers have expanded their systems to include real-time translation functions. Google made the first significant strides with new Pixel devices and Gemini-powered translation features, whilst Apple also began rolling out Live Translation in combination with Apple Intelligence. At the same time, increasingly specialised hardware solutions are appearing: translation earbuds or AI-enabled headphones that simultaneously translate conversations between multiple languages whilst working directly with a smartphone or an on-device model. This development shows that by 2025, real-time translation will no longer be purely a software function, but will increasingly be integrated into wearables and smart devices.
However, what has not yet become widespread
Despite this progress, usage remains limited in many areas. Many features are rolled out regionally or only on specific devices, are restricted to certain languages, or depend on new hardware. Quality varies depending on the ecosystem, and complex translations involving subtle nuances or cultural contexts continue to pose a challenge. For businesses, this means: the technology is becoming more suitable for everyday use, but is not yet ubiquitous.
Context
Multilingual real-time communication has taken a significant step forward in 2025 and has moved closer to everyday use – not least due to its integration into smartphones, earbuds and other IoT devices. However, the major breakthrough is not expected until 2026, when generative AI capabilities become more widely available and the hardware infrastructure is rolled out across more regions and device ranges.
Ethical AI and Governance
In our annual outlook, we had anticipated that the EU AI Act would become a key framework for the responsible use of AI by 2025 – with clear requirements regarding transparency, security and the protection of fundamental rights. Looking back, it is clear that this assessment was correct. With the August 2025 milestone, key parts of the legislation came into force, particularly in the areas of governance and obligations for general-purpose AI. The regulation is thus taking on a clearer shape, even if not all detailed issues have yet been definitively resolved.
What was actually achieved in 2025
With the entry into force of the first prohibitions under Article 5 of the AI Act in February 2025, several high-risk applications were explicitly banned – including social scoring, certain forms of predictive policing, emotion recognition in the workplace and education, and the mass scraping of facial images. In parallel, governance structures were established at EU and national level, such as the European AI Office and the relevant supervisory authorities in the Member States. Providers of general-purpose AI models are also facing stricter transparency and documentation requirements, the implementation of which was the subject of intensive preparations in 2025.
What practice shows
For businesses, 2025 means, above all: taking stock, assessing and documenting. AI systems must be categorised, risks assessed and processes adapted – an approach which, as is usual with major EU regulatory frameworks, involves considerable organisational effort and raises questions of interpretation. This applies particularly to areas such as emotion recognition, social scoring or complex agent systems, where the guidelines have been defined but have not yet been operationalised in every detail. This is precisely why it is so valuable to have a competent and technologically experienced IT partner at your side during the transformation process, one who supports companies in establishing sustainable AI governance structures.
Context
Our forecast that ethical AI and governance will evolve from a ‘nice-to-have’ to a key location factor by 2025 has thus been clearly confirmed. The EU AI Act has become part of day-to-day operations and is shaping the way in which AI is developed, assessed and deployed in Europe. The resulting complexity is high – yet those who establish robust processes at an early stage will secure long-term legal certainty, strengthen the trust of their users and position themselves for the future in the European market.
Conclusion and Outlook
Looking ahead to 2025, we see a year in which conversational AI has become significantly more mature – less through disruptive leaps and more through consistent technical progress, more stable architectures and clearer regulatory frameworks. Many of the trends that still seemed visionary at the start of the year have now taken concrete shape, even if widespread implementation will only become apparent in some areas as late as 2026. With the increasing availability of generative AI across platforms, devices and applications, and with regulation creating reliable guidelines, an environment is emerging in which AI-supported interaction is becoming an integral part of digital services. Companies that continue to invest in technologies, governance and expertise now will secure a sustainable competitive edge – and will be well-positioned to play an active role in shaping the next phase of conversational AI’s development.
