AI adoption is not straightforward for corporates

Most companies (especially small and medium companies, SMEs) are not set up to develop, test, maintain and upgrade in-house software using AI tools. We also believe that, because SME customers prioritise accuracy, predictability and auditability of the software they use, they will favour AI-augmented products from existing trusted partners (such as HgT’s portfolio companies) underpinned by extensive domain expertise over cheaper but more experimental AI-driven solutions.

Hg is reporting good traction with AI projects

While Hg (HgT’s investment manager) observes that AI adoption among enterprises has resulted in softer growth of traditional software-as-a-service (SaaS) revenues, it has also experienced a strong pick-up in bookings for AI-enabled products offered by its portfolio companies. New product innovation is accelerated by its dedicated AI product incubator (Hg Catalyst). Hg expects the over 700 live generative AI projects across its portfolio to deliver c $130m of incremental EBITDA this year, up from c $40m last year. We consider AI’s impact on software pricing as a key uncertainty, though Hg recently noted that initial data from its portfolio suggests that margins from AI ‘products of action’ are similar to traditional SaaS ‘products of record’.

Hurdles to challenging SaaS incumbents

We believe that there are multiple potential hurdles for AI-powered in-house development of software by corporates and AI start-up challengers trying to disrupt incumbent players, such as HgT’s portfolio companies. Most companies are not set up to develop, test, maintain and upgrade in-house software. The long-term trend has been for corporates to develop less in-house software and focus on the core business. This especially makes sense for mission-critical solutions in clusters such as accounting, enterprise resource planning (ERP) and payroll (offered by HgT’s portfolio companies), where SME customers are interested first and foremost in the software’s accuracy, predictability, regulatory compliance, control and auditability of outcomes when handling large volumes of transactions/records in deterministic workflows. This may prove difficult to facilitate using generative AI (GenAI) on a standalone basis given the probabilistic (rather than deterministic) nature of its output. The complexity of edge cases as well as the fragmentation of the European regulatory framework (74% of HgT’s portfolio companies at end-September 2025 are Europe-domiciled) cannot be easily addressed with innovative technology alone. This is because it requires deep vertical-specific and regional domain expertise acquired over many years, in terms of keeping up with legislative change, encoding it safely, proving it in audits and handling exceptions at scale. Products deeply embedded in the client’s system, combining a system of record, workflow and audit trails, cannot be replaced effortlessly. Corporates use multiple pieces of software (sometimes hundreds of them) and all of them need to be integrated into existing (often complex) workflows.

Furthermore, these workflows are built on proprietary datasets such as tax filings, regulatory records and transaction histories, which companies may be reluctant to share with generic AI tools offered by the main large language model (LLM) providers or AI startups for security and business confidentiality reasons. Software providers that are also providers of data powering the software can restrict access to potential challengers.

Therefore, it seems more likely that AI agents will be used to handle parts of these workflows (under the oversight of human experts), especially non-critical and decision-support functions, rather than replace traditional software completely.
For the above reasons, we believe that many SME customers will favour AI-augmented products from existing trusted partners underpinned by strong domain knowledge over cheaper, but more experimental AI-driven solutions.
Hg believes that AI adoption will be driven by incumbent software providers catering to the needs of regulated sectors, professional-led advisory and European SMEs that do not have huge IT budgets. In order to succeed, HgT’s portfolio companies have to successfully build innovative AI-driven ‘products of action’ on top of the base layer and wrap them into appropriate controls (structured outputs, rules, verification, human-in-the-loop setups and audit logs). Otherwise, challengers will offer such solutions on top of the base software offered by the traditional SaaS provider.

Acting as an AI disruptor for its own portfolio companies

While there is always the risk that incumbents will suffer from being behind the curve in terms of AI adoption, we noted in our November review note that Hg has been an early adopter exploring the potential of AI and has built a team of 20 in-house data and AI experts and more than 60 consultant specialists, whose experience has been codified in Retina, a suite of AI-powered SaaS tools available to Hg’s portfolio companies. Retina augments the business intelligence, data-driven decision-making and M&A sourcing of HgT’s portfolio companies.

Furthermore, Hg is collaborating closely with several of its portfolio companies to transition to an AI-first business model.
Given that during a platform shift like AI it is difficult to predict which approaches to product development will work best, Hg encourages its portfolio companies to experiment (while setting clear boundaries, eg on the legal side) and to collaborate with other portfolio companies to share their ideas and experience with AI projects. It recently launched Hg Catalyst, its dedicated AI product incubator to accelerate product innovation across its portfolio. Hg Catalyst leverages more than 80 AI engineers, designers and product managers who work directly with Hg’s portfolio companies to build agentic AI. The incubator focuses on defining narrow workflows with clear success metrics, fast release cycles and the reuse of best practices (to facilitate the rollout of new products in weeks based on experience gained from previously launched products).

Hg’s capabilities and value creation activities across its portfolio related to AI solutions should give its holdings a competitive advantage over peers that do not have such strong backing. In fact, we believe that the AI threat could lead to a stronger flow of add-on opportunities for Hg as smaller software companies that struggle to roll out agentic AI tools decide to obtain the backing of an experienced software investor like Hg. This could lead to a stronger positive impact of multiples arbitrage from Hg’s ‘buy-and-build’ strategy, especially in the fragmented European software market, in which Hg mostly operates.

Top-line growth to be increasingly driven by AI adoption

Average top-line growth across HgT’s portfolio was 17% in FY25, somewhat softer than the 2020–24 average of 25% for its top 20 holdings. Two points need to be considered in this respect. First, the above-mentioned five-year period covered the COVID-19 pandemic (during which many companies pulled forward software spend) and one-off, inflation-driven price increases in 2022 and 2023. Second, 2025 growth was affected by the tariff-induced ‘uncertainty pause’ in global IT spending in H125.

That said, Matthew Brockman (chair of Hg’s investment committee) recently commented that traditional SaaS revenue growth rates are slowing, driven by a pick-up in AI spend from end-customers. Hg remains confident in the potential of AI-enabled products developed by its portfolio companies, indicating that they are rapidly building their share in new business pipelines (up to 40% of new bookings), at the expense of growth in seats, as well as upsell and cross-sell activities. Therefore, Hg hopes to see a more meaningful impact on revenue and margins across HgT’s portfolio in the next one to two years, with over 700 live GenAI projects (including over 40 AI-enabled product features that have already been launched) expected to deliver c $130m of incremental EBITDA this year, up from c $40m last year.

Hg believes that agentic AI could significantly expand the addressable market for products of its portfolio companies (even by 10 to 15 times) by taking on more of the processes currently handled by white collar workers and facilitating lower delivery friction (thereby making it economically viable to offer products to new customer groups). According to major private tech investors, AI provides an opportunity to go beyond the traditional rule of 40, which is a rule of thumb (but based on empirical evidence) saying that software stocks with combined revenue growth and EBITDA margins of 40% or more typically generate higher valuations. According to the Global Private Equity Report 2025 published by Bain & Co, Vista Equity Partners (a private equity investor in enterprise software, data and technology companies) believes that AI will rewrite the rule of 40 and drive the combined revenue growth and margin of SaaS businesses to 50% or even 60%. Orlando Bravo (the founder of Thoma Bravo) shared similar views during a recent CNBC interview. Nic Humphries, Hg’s senior partner and executive chairperson, expects the combined revenue growth and EBITDA margin of SaaS businesses that successfully develop AI-powered products to increase by 5–10pp.

AI’s impact on pricing and margins the key risk factor

We acknowledge that, even if the ongoing AI disruption does not shrink the client base of traditional SaaS companies, the AI-enabled higher productivity per employee could put pressure on software prices or at least create uncertainty with respect to the future pricing model (which at present is normally based on a ‘per seat’ formula).

Pricing pressure could be at least partly offset by the productivity gains realised by software companies from using AI for product development (code generation and modernisation in particular), as well as customer support. Based on our discussion with Hg, we understand that R&D teams across its portfolio companies are achieving engineering productivity improvements of two to 10 times using agentic coding assistants. Moreover, up to 50% of customer support cases are resolved end-to-end by AI (with most advanced companies achieving an up to 90% automation rate). This reduces cost and improves customer satisfaction scores, and also leaves the most complex tasks to employees and in turn leads to a reduction in staff churn. So far, margins seem to be holding up well, with HgT reporting an average 34% EBITDA margin across its portfolio in 2025 (vs a 2020–24 average for its top 20 holdings of 32%), coupled with an average 20% y-o-y EBITDA growth.

Hg recently highlighted that initial data from its portfolio suggests that margins from AI ‘products of action’ are similar to traditional SaaS ‘products of record’. This is because purely consumption-based pricing models are not proving popular given the associated spend uncertainty (we believe that they may also prove difficult to implement). However, we cannot rule out the emergence of hybrid models with a base fee and a consumption-based component, which could prove to be more popular among customers.

Last but not least, software pricing considerations also need to account for the inference cost of LLM models. Major LLM providers seem to have been heavily subsidising their products to facilitate growth. However, when pressure starts to mount on these companies to demonstrate a clear path to profitability in light of the considerable capital expenditures on infrastructure (including data centres), the price per token could go up, reducing the competitive advantage of in-house software built using AI (although this may also lead to a higher marginal cost of providing AI-augmented software per user by traditional SaaS companies).

Discount to NAV offers substantial safety margin

We acknowledge that the impact of AI on SaaS is difficult to predict at this stage. However, HgT’s current discount to NAV offers a certain safety margin for investors. At the last closing price of 455.5p, HgT’s shares trade at a 19% discount to the estimated NAV of 561.9p as at end-December 2025. This is significantly greater than the average 10-year discount of 7%(see Exhibit 1). We note that all HgT’s non-executive directors bought HgT shares recently, either personally or via a person closely associated.

HgT values most of its holdings using last-12-month (LTM) earnings (or forward earnings if they are expected to be lower than LTM earnings). The multiples HgT uses to value its holdings are a blend of comparable listed businesses and, where relevant and recent, private M&A deal multiples. The latter tend to be less volatile than listed multiples and have a smoothing effect against public multiples because they are not subject to the same short-term swings in investor sentiment.

While short-term and cyclical changes in public software multiples affect HgT’s portfolio carrying values (the company stated that roughly 20–40% of public comparable movements flow into its valuations), we note that its long-term returns have historically been driven primarily by the strong earnings growth of its portfolio companies, not multiples expansion or contraction. In this context, we note that HgT has historically delivered revenue and EBITDA growth of around 20–30% per year (c 10–15% is organic growth) at a margin above 25% across its top 20 holdings (see Exhibit 3 above).

The ultimate validation of HgT’s portfolio valuations are exits at or above last carrying value, and HgT has a strong track record in this respect. Between 2017 and 2022, the trust’s exits were completed at an average uplift to carrying value at the end of the preceding year of a solid 38%. Even in the more demanding environment after 2021, marked by low deal activity and pressure from higher interest rates, the trust was able to realise investments at an uplift. In 2024, this stood at an average of 15% compared to the quarter before signing the transaction. This brought its average uplift in 2020–24 to 22% versus the carrying value at point of signing. Its 2025 exits were completed at an average uplift of 25% and resulted in an NAV accretion of 4.6p to HgT (0.8% of its opening NAV). We believe that all the above is a testament to the quality of the trust’s portfolio and its conservative valuation approach.

Of particular note are the two recent exits of GTreasury (announced in October 2025) and Intelerad (announced in November 2025), both to strategic buyers, at uplifts to carrying value of 97% and 62%, respectively. We understand that the companies’ AI product roadmaps were a key diligence item for the buyers. This may be an indication that pricing in the private markets has already begun to differentiate between the ‘AI winners’ and the ‘AI losers’, despite the indiscriminate sell-off in public SaaS companies.