Latest Trends in GenAI, 2025 and Beyond: Research Shows the Pros and Cons
The rise of generative AI shoots higher than the Old Faithful geyser in Yellowstone National Park since ChatGPT dropped into our laptops in November 2022.
With promises of redesigning workflows to boost productivity, drive revenue growth, and increase profitability, venture capital companies pumped $56 billion into GenAI in 2024 – nearly tripling 2023 figures.
But how is that investment trickling down into the real world, or more specifically, is it solving business-critical challenges as anticipated?
This was one of the burning questions ABBYY sought to answer by commissioning its eighth global annual State of Intelligent Automation Survey to uncover the good, the bad, and the risks of GenAI.
Here, I break down the results of the Generative AI Confessions Report 2025* to reveal leaders’ experience with the technology, how it impacted business, and plans for the future.
GenAI deployment isn’t easy
It was reassuring to see that an incredible 98% of respondents to the survey were happy with GenAI results overall. However, this came with a giant BUT, as many faced extensive hurdles during implementation.
Unfortunately, the number one problem was integrating GenAI into current business processes, with more than a quarter (27%) saying this was a major challenge. The same number said training models was harder than expected, while 24% of US managers reported that employees misused the tools — the highest globally.
In addition, 22% did not have proper governance, and a fifth (20%) had problems with hallucinations.
Sadly, as a result of these complications, an incredible 23% of US companies decided to stop using the tech altogether. This defeatist approach by US businesses was much higher than other countries, with only 8% of French companies and 10% of UK leaders opting to abandon the deployment rather than fix the issues.
So, for those who decided to stick it out, how did they resolve the problems?
GenAI needs a little help from a friend
We’re all familiar with the saying “one-size-fits-all,” but let’s face it, that’s rarely true. And the same goes for using GenAI— it’s not a magic bullet.
Respondents to the survey reported needing to augment GenAI with other technology in order to improve outputs and get the results they wanted.
In the US, 40% added AI agents, over a third (36%) turned to process intelligence, while 31% introduced document AI/IDP. Retrieval augmented generation (RAG) was also leveraged to improve success by 23% of managers.
Honestly, this didn’t come as a surprise at all. It makes total sense that organizations would turn to process intelligence to create a visibility map of their workflow in order to get to the root of what is going wrong and where GenAI can best help. In fact, it would have been wise to evaluate their processes BEFORE investing in GenAI. They may have even discovered they didn’t need GenAI in the first place, and that other, more purpose-built AI would suffice.
It’s a similar situation with RAG, which can be combined with LLMs and other technologies for anything from improving chatbots and virtual assistants to troubleshooting technical issues to generating research materials and reports. Integrating RAG with IDP leads to more accurate and relevant responses, improvements in data management, including data scrubbing, redaction, and normalization, before it is fed into LLMs. This keeps the data as accurate and relevant as possible. The quality of the IDP data directly impacts the LLM’s ability to generate accurate, relevant, and coherent text.
As the survey showed, taking this augmented approach with GenAI led to automation projects being far more successful. Respondents reported better consistency of outputs (58%), better integration into existing workflows (50%), more accurate and reliable results (48%), greater cost efficiency and savings (44%) and increased user trust (42%). I rest my case.
GenAI is sparking shadow AI
Perhaps the fact that ChatGPT is being used by up to 79% of workers is a good indication that GenAI investment is being pushed from the ground up.
The survey saw company leaders admit that the main reason they invested in GenAI is that employees are already using it. In fact, the only reason GenAI was operational in a fifth of businesses (22%) was that staff were installing it on a bring-your-own-software (BYOS) basis rather than as part of a management-driven decision.
It’s no shock then that almost half of leaders (43%) cite staff usage as a trigger for strategic deployment.
What is concerning, however, is the amount of shadow IT that GenAI is causing, which can potentially raise serious data privacy concerns. Not having a uniform overview of which software and licenses are in circulation can inadvertently lead to compliance issues. The more unknown applications are used within a company, the higher the risk that outsiders can gain access to the entire corporate network and increase the potential for a cyber-attack. Obviously, the IT department cannot protect itself from threats it does not know exist. Worryingly, the survey showed it is the US that is seeing some of the highest rates of BYOS, with 22% of companies experiencing it, compared to 12% in the UK and 17% in Australia. Only Singapore was higher at 26%.
One relevant trigger for shadow AI growth in recent years has been the new generation of digital natives — those born into technology now entering the job market and bringing with them a high level of digital affinity. When they encounter digitalization hurdles, they want to take things into their own hands and download solutions to their problems or independently make their working processes more efficient.
There is a clear appeal for staff in wanting to install GenAI through the back door. Our survey showed that over half (55%) said it reduces workload, increases their productivity, and supports creativity and innovation. Interestingly, 53% admitted they use it to make themselves look smarter and more professional. It was reassuring to see that almost half (48%) want to develop and expand their use of the tools. Overall, the feedback from employees is good, with 91% having a positive experience with GenAI tools.
GenAI is top dog among AI technologies
While GenAI might have its challenges in terms of deployment, it remains one of the most popular AI technologies among US companies.
An incredible 86% of business leaders said they are currently using GenAI in their organization, compared to 65% using purpose-built AI, and 55% utilizing agentic AI. There’s no getting away from the fact that leaders feel GenAI is worth the money, but what are they using it for? Here’s what they told us:
- Data analysis and insights generation 64%
- Employee productivity (writing assistance/project management) 59%
- Risk management and compliance 53%
- Customer service operations 51%
- Automating document business process (such as AP and logistics) 48%
GenAI comes with risks
The problem of hallucinations with GenAI and LLMs is an ongoing issue that has had far-reaching consequences. Take the case of the lawyer fined for submitting a legal brief containing fake citations that were generated by ChatGPT. LLMs’ inclination to lie is particularly worrying given that AI hallucinations have DOUBLED in the last year.
While our survey showed that a fifth of US companies experienced hallucinations when implementing GenAI tools, it was also interesting to see that employees themselves don’t trust it. When asked what improvements they’d like to see with GenAI, 23% of US workers wished that it would stop lying and admit when it didn’t know the answer. Concern about GenAI not being truthful is highest in the US, with only 13% of workers in Germany and 15% in the UK feeling the same.
The stakes are high, however, for business leaders when GenAI does not deliver consistent, reliable results, as it can impact compliance, reputation, and spark legal issues. AI systems learn from data, and biased data leads to biased decisions. Data readiness plays a fundamental role in achieving reliable AI outcomes. Yet, its importance, especially among non-technical leadership, is often underestimated.
This is why it is crucial to address foundational issues like data quality and reliability to ensure technologies like GenAI can deliver on their promises. It makes total sense that the respondents in the survey turned to other tools like IDP and RAG to improve data excellence and output.
Technologies fostering trust and mitigating risk are essential, particularly as we deploy more advanced systems across our operations. The survey confirmed the problems organizations are facing with GenAI compliance, with 26% of global respondents not having proper AI policies in place to deploy it.
GenAI investment will continue — but are we doing enough?
Despite the challenges, there is definitely no slowdown in investment in GenAI.
Gartner reports that overall global AI spending is expected to exceed $2 trillion in 2026, led by GenAI being integrated into products such as PCs and smartphones, as well as infrastructure.
Our survey also reflected an upward trajectory in AI investment with an average increase of 21% globally, while 1 in 9 will plan on raising budgets by a whopping 50% or higher. However, the US may need to step up its game to keep pace with its European and APAC counterparts. Business leaders in the US reported spending an average of $794K on AI tools in the last 12 months, lagging well behind Singapore ($1.4M), France ($1.2M), the UK ($1.1M), Germany ($1M), and even Australia ($895K).
It’s easy to see that GenAI is bringing enormous value to organizations worldwide by automating tasks, fueling creativity and optimizing costs. However, its potential for harm by offering misinformation is significant, and this must be addressed by business leaders through strategic adoption with risk management as a priority
*Opinium Research conducted interviews with 1,200 senior managers or above in companies of 100+ employees (most with over 1,000) in the US, UK, France, Germany, Australia and Singapore.
Maxime Vermeir
Maxime Vermeir is Senior Director of AI Strategy at ABBYY.