Nuix (NXLLF), Tongdao Liepin (TGDLF) Price AI as a Separate Tier

On 23 August 2026, Nuix Limited (NXLLF) and Tongdao Liepin Group (TGDLF) held separate earnings calls on the same day, and both described the same change: AI has stopped being a feature bundled inside the licence and is now a separately priced tier sold to customers they already had. Neither company sells a model. What each one prices is governed access to a body of data the buyer cannot assemble anywhere else.

---

What these companies sell, and what changed

Nuix makes forensic and eDiscovery software. Companies, law firms and law-enforcement agencies use it to turn raw material — email, disk images, chat logs — into evidence that will hold up in court. Its platform product is Neo, with 135 paying customers [1]. Tongdao Liepin runs a mid-to-senior recruiting marketplace in China, where employers pay to post roles and reach candidates: 1.51 million registered business users and roughly 63,000 that actually pay [2]. Historically both companies folded new capability into the existing licence or membership fee, and AI arrived the same way.

What changed is the object being charged for. General-purpose models such as GPT and Claude are now interchangeable for most tasks, and buyers who have to account for their process in court or to a regulator have started insisting on running the specific model they chose against data they cannot get elsewhere. Nuix productised that as BYO AI: the customer brings its own model, and Nuix supplies the curated, defensible case data [1]. The model is the customer's choice and the inference bill lands on the customer's side, so what Nuix sells is the governed access itself — which can be lifted out of the licence and sold again, at a higher price, stacked on the existing contract. The binding condition is not who pays for inference. Tongdao Liepin runs its own vertical model, Tongdao Hui Cai, carries that cost itself, and its gross margin still rose [2]. What matters is whether the buyer can assemble that data somewhere else.

---

Attach rates, revenue per customer and gross margin

Nuix disclosed attach and retention. About 60% of its 135 Neo customers have bought the AI-enriched solutions library and about 20% have gone on to buy the higher AI enablers — BYO AI, Semantic Search and Transcription. Net dollar retention finished the year at 105.2%, up from 101% at the December half, which the company attributed to focused upsell [1]. An analyst put the obvious question directly to management: does that uplift bring pass-through inference costs and compress margin? CEO John Ruthven said no, and added that margins are holding around historical levels [1].

Tongdao Liepin disclosed coverage and a price ladder. AI-account coverage has reached nearly 100% of its paying base. Within six months of the AI Pro account launching, around 40% of AI-account customers had upgraded and more than 70% of new customers chose AI Pro outright. First-half cash billings from AI products grew nearly 100% year on year while the paid customer count rose 0.4%, and gross margin improved 1.4 percentage points to 79%, which the company attributed to a better product mix as AI revenue contribution rose and to higher delivery-team productivity [2].

The same research contains a case that runs the other way. CS Disco (LAW), which competes directly with Nuix Discover in eDiscovery, reports more matters and larger matters, attributing that to simpler pricing and to its Cecilia AI assistant being included on every matter; its gross margin was 76%, flat against 76% a year earlier [3]. What Disco sells as AI is model capability rather than data the buyer cannot replace, and the same mechanism pushes it to give AI away to win deal flow instead.

---

The control point moves from model capability to data access

If the priced object really has moved to data access, the operating metrics of these companies change shape. New revenue comes from customers already under contract, so the customer entry point does not get reallocated and the logo count can stay flat for a long time. Pricing shifts from a one-time renewal increase to a ladder that can be stacked repeatedly. Growth therefore shows up in retention and revenue per customer: Tongdao Liepin's paid customers up 0.4% against AI cash billings roughly doubling is the plainest reading of that [2].

The cost side needs more care. Nuix pushed the inference bill onto the buyer and Tongdao Liepin absorbed it, and neither margin deteriorated — which points to the pricing power of the data rather than to where the inference bill sits [1][2]. Two figures are worth checking from here: whether attach at the higher tier keeps climbing (about 20% at Nuix, about 40% at Tongdao Liepin), and whether gross margin still holds or rises in the next reporting period. One precondition is still unmet. Neither company has published a revenue figure for this AI layer, and Nuix said explicitly that it does not have enough track record to put a number in the market [1]. Until it does, no outsider can independently size the layer's contribution to growth.

---

Companies exposed to the same change

• Open Text (OTEX): Runs enterprise content management, holding customers' contracts, email and records. It sells Aviator agents inside existing portfolio deals, says deal size is four times larger when Aviator agents are included, and reported non-GAAP gross margin up 220 basis points to 78.3%, attributed to lower hyperscaler costs [4]. Same point in the chain, but it has not disclosed an installed-base attach figure.
• CCC Intelligent Solutions (CCC): Operates the US auto-insurance claims data network, and packages AI as separately purchasable modules that existing customers add at renewal. AI-based solutions are around 11% of revenue growing roughly 45%, with net dollar retention of 107%. It is also the boundary case: it owns its model layer, and adjusted gross margin fell from 78% to 76% over the year [5].
• John Wiley & Sons (WLY): An academic publisher, and an upstream source of exactly the kind of data buyers cannot assemble themselves. It already packages content into AI subscription knowledge feeds sold directly to 19 corporate customers, up from 10 the prior quarter [6]. If the priced object really is data access, content owners sit further up this chain.

---

For owners of hard-to-replace data, this is an option rather than an automatic outcome. Kanzhun (BZ), which holds a comparably large data set in Chinese recruiting, had its CEO say on the May 2026 call that he is choosing to grow the paying customer count first and leave price increases for later [7]. Cellebrite (CLBT), a peer in investigations software, disclosed that the price uplift it captures at each Insights conversion is getting smaller [8].

---

Sources

[1] Drillr · Nuix Limited (NXLLF) · 2026-08-23 · FY2026 full-year results call

No. I mean generally, the uplift, the same level of margins applying.

[2] Drillr · Tongdao Liepin Group (TGDLF) · 2026-08-23 · 1H 2026 results call

[3] Drillr · CS Disco (LAW) · 2026-08-05 · Q2 2026 earnings call

[4] Drillr · Open Text Corporation (OTEX) · 2026-08-06 · Q4 FY2026 earnings call

[5] Drillr · CCC Intelligent Solutions (CCC) · 2026-07-30 · Q2 2026 earnings call

[6] Drillr · John Wiley & Sons (WLY) · 2026-06-16 · Q4 and full-year FY2026 earnings call

[7] Drillr · Kanzhun Limited (BZ) · 2026-05-20 · Q1 2026 earnings call

[8] Drillr · Cellebrite (CLBT) · 2026-08-13 · Q2 2026 earnings call

This is only meant to surface industry changes and companies you may have overlooked - it is not a stock recommendation.