# ai_software.md — AI Software Agent

**Bucket:** AI Software & Models (early monetisation)
**Mission:** Summarise what investors, analysts and fund managers are watching in enterprise AI,
model providers, automation platforms and the data layer.
**Last reviewed:** 1 August 2026

> **This is market research, not financial advice.** Nothing here is a recommendation to buy or sell
> anything. Every opinion below is attributed to the party that expressed it. Company names appear as
> examples of where a market narrative is being expressed, not as endorsements. No live price or
> valuation data is used — verify everything independently.

---

## 1. Agent mission

Track the layer that has to eventually justify the infrastructure spend. If the AI Infrastructure
bucket is the railroad, this bucket is the freight — and the central question every analyst in this
space is now asking is whether the freight is actually showing up, and at what price.

2026 is the year that question stopped being rhetorical. The framing across analyst notes has shifted
from *adoption* to **monetisation**: not "are enterprises using this" but "are they paying, how much,
and does it renew."

---

## 2. Bucket analysis

### 2.1 The contrarian fact that frames this entire bucket

Per Goldman Sachs data on Q2 2026 hedge-fund positioning, **software fell to roughly 6% of hedge-fund
long portfolios — the lowest weight since 2019** — at the same time semiconductors hit a record ~10%.

Funds have been expressing the AI trade through hardware, not software.

This cuts two ways and honest analysis has to hold both:

- **The bearish read:** professional money doesn't believe software captures the value. The concern
  is that AI compresses the price of software rather than expanding it — if an agent can do the work,
  the seat-based licence that used to charge for a human doing it is at risk.
- **The contrarian read:** by this site's own hype-cycle framework, a nine-year-low positioning
  weight is the *opposite* of crowding. Emergent stages are defined by cautious sentiment and thin
  participation, which is exactly what a 6% weight describes.

**This is the single most interesting structural fact across all three buckets** — the theme with the
most attention (infrastructure) and the theme with the least positioning (software) sit on opposite
ends of the same trade.

### 2.2 The monetisation evidence

The data investors point to when arguing revenue is genuinely arriving:

- **OpenAI reports over 40% of revenue now comes from enterprise clients** — a shift from consumer
  subscriptions toward contracted business spend.
- **Zylo's 2026 SaaS Management Index** shows AI-native application spend **up 108% year over year**.
- The SaaS delivery model for agentic AI is described as growing at a **46.8% CAGR**.
- Sapphire Ventures frames 2026 as "software's AI inflection point"; Vista Equity Partners has
  published on software's transition to agentic enterprise AI.
- A review of ~40 SaaS earnings calls concluded AI will be the biggest boon to the space — notable
  because it is drawn from company disclosure rather than forecast.

### 2.3 The pricing-model shift — the mechanism that matters

The most analytically useful narrative in this bucket: pricing is moving from **"pay for access" to
"pay for work done."**

This is not a marketing distinction. Seat-based SaaS pricing caps revenue at the number of humans a
customer employs. Outcome or consumption pricing uncaps it — but only for vendors who can *measure*
the work an agent performed and bill for it.

Commentary names **Salesforce, ServiceNow, Intercom and HubSpot** as vendors demonstrating that AI
agents can expand the total addressable market rather than cannibalise it. The claim analysts make is
that companies which solve the measurement-and-billing problem capture the next wave of growth.

**The test this implies** is concrete and checkable: in the next set of results, is AI revenue
disclosed as a *separate, growing line item* with a retention figure attached — or is it described
qualitatively in the CEO letter? That distinction separates the two halves of this bucket more
cleanly than any valuation metric.

### 2.4 The counter-narrative: the ROI reckoning

The bear case is not fringe and deserves equal weight:

- Enterprise AI is described as hitting an inflection point **as companies rein in spending and
  demand real results**.
- Commentary references an enterprise AI spending "ROI crisis," with very large aggregate spend
  figures cited against thin demonstrated returns.
- Procurement is reportedly getting more stringent — pilots that ran on innovation budgets are now
  being asked to justify renewal from operating budgets.

Both things are happening simultaneously: **AI-native spend up 108%, and buyers getting harder to
sell to.** The reconciliation most analysts reach is consolidation — spend growing overall but
concentrating into fewer vendors who can prove outcomes. If that is right, this bucket has a
widening gap between winners and losers rather than a rising tide, which makes it a
*discrimination* trade rather than a *theme* trade.

### 2.5 The layers investors segment by

| Layer | What it does | Names analysts commonly cite |
|---|---|---|
| **Model providers** | Frontier models sold via API and enterprise contracts | OpenAI, Anthropic (both private) |
| **Data layer** | Where enterprise data lives and is queried | Snowflake, Databricks (private), MongoDB |
| **Applied AI / decision platforms** | Deployed analysis and operations | Palantir |
| **Agentic workflow platforms** | Software that performs work, not just displays it | Salesforce, ServiceNow, HubSpot, Intercom |
| **Observability & tooling** | Monitoring the systems built on all of the above | Datadog and peers |

A structural note investors raise often: **the two most important model providers are private.** The
purest exposure to this bucket cannot be bought on an exchange, which pushes public-market money into
adjacent layers and arguably inflates them relative to the value actually being created.

### 2.6 The two named comparisons

**Palantir (PLTR)** — the most-discussed name in the bucket. Analysts expect earnings growth of
**~95% in 2026 followed by ~42% in 2027**, against a stock trading around **65x forward earnings**.
Wall Street consensus is a **"Moderate Buy," with 19 of 28 analysts at "Strong Buy."** The bull case
is defence and government contracts plus genuine deployed-AI revenue; the bear case is that the
valuation already prices several years of that growth, leaving no room for a miss.

**Snowflake (SNOW)** — expanded a **$6bn, five-year agreement with AWS** focused on generative and
agentic AI infrastructure and migrating AI workloads onto its platform. Analysts describe the data
layer as strategically essential because AI is only as useful as the data it can reach.

**The structural contrast** analysts draw between them is genuinely useful: Palantir's model involves
significant upfront investment and customisation, while Snowflake and Databricks run
**consumption-based** models that scale with usage. Consumption models are more sensitive to a
spending slowdown but scale faster in an upswing — the same business in different market conditions
produces very different results.

### 2.7 ASX exposure — thin, and worth saying plainly

The ASX has little genuine AI-software exposure. Names that appear in Australian AI baskets —
**WiseTech Global (ASX:WTC)**, **TechnologyOne (ASX:TNE)** — are quality software businesses whose AI
linkage is more *applied* than *native*. Anyone wanting real exposure to this bucket is looking
offshore. Forcing an ASX name into it would be a worse answer than admitting the gap.

---

## 3. Must-watch early plays — as investors describe them

**"The monetisation inflection."** Wedbush describes 2026 as the third year of a ten-year AI cycle
and expects "transformational" monetisation opportunities, with US big tech as main beneficiaries.
The narrative: infrastructure was built in years 1–2, revenue arrives in years 3–5.

**"Pay for work done."** The pricing-model shift in §2.3, treated as the key variable. Investors
applying it look for vendors disclosing consumption or outcome-based AI revenue as a separate line.

**"The data layer is the toll booth."** The argument that models commoditise but the enterprise data
estate does not — whoever stores and governs the data captures durable value regardless of which
model wins. Snowflake's AWS expansion is the reference point; Databricks is the private comparison.

**"Applied beats general."** The Palantir framing: value accrues to whoever deploys AI into a
specific workflow with a contract attached, not to whoever has the best benchmark score.

**"Software is under-owned."** The genuinely contrarian one, and the most interesting for anyone
hunting Emergent-stage narratives. If a 6% hedge-fund weight is a nine-year low while AI-native
spend grows 108%, positioning and fundamentals are pointing in opposite directions. Investors making
this argument treat the gap as the opportunity. It requires being early and being wrong for a while
first — which, as the main page notes, is often indistinguishable from just being wrong.

---

## 4. Risk tiering — this bucket

### Low risk (relative)

- **Large-cap platform incumbents** — Microsoft, Alphabet, Amazon, Salesforce. Analysts describe
  these as the lowest-risk expression: AI is upside on an already-profitable base rather than the
  whole thesis. Wedbush's view of big tech as the main monetisation beneficiary sits here.
- **Established data-layer names** — Snowflake, MongoDB. Real recurring revenue and strategic
  positioning; growth expectations are demanding.

### Medium risk

- **Palantir** — genuine revenue and contracted government work, but ~65x forward earnings on
  expected 95% then 42% growth means the risk is almost entirely in the multiple rather than the
  business.
- **Agentic workflow platforms** — ServiceNow, HubSpot and peers. Real products shipping; the open
  question is whether agents expand the seat model or erode it, and the answer is not yet in the
  numbers.
- **Observability and tooling** — Datadog and peers. Benefits from complexity, exposed to consumption
  slowdown.

### High risk

- **Pure-play AI application companies with thin revenue** — where the ROI reckoning bites hardest.
  Commentary suggests procurement is now selecting against exactly this profile.
- **Anything priced on AI narrative without disclosed AI revenue.** The hedge-fund test from the
  infrastructure bucket applies identically: is the exposure promotional or economic?
- **Private model providers accessed via secondaries or feeder structures.** Access is the whole
  game, fee layers are heavy, and pricing is opaque.

---

## 5. Risks to the whole bucket

1. **The ROI reckoning resolving badly.** If enterprises conclude AI pilots didn't pay, the spend
   growth reverses fast — software budgets are more cancellable than data-centre contracts.
2. **Price compression.** The deepest structural risk: AI may make software cheaper to build,
   inviting competition into every category simultaneously.
3. **Model commoditisation cutting both ways.** Good for application companies (cheaper inputs), bad
   for anyone whose thesis is model superiority.
4. **The best assets are private.** Public-market exposure is a proxy, and proxies drift.
5. **Consumption models cut in reverse.** The same usage-based pricing that scales beautifully upward
   de-scales just as quickly when customers optimise spend.
6. **Positioning risk in reverse.** A 6% hedge-fund weight is thin support. Under-owned assets fall
   further on bad news precisely because there is no marginal buyer.

---

## 6. Conclusion

This bucket is the mirror image of AI Infrastructure, and that is what makes it interesting.

Infrastructure has record hedge-fund positioning and a narrative the main page classifies as
Saturation-to-Decay. Software has its lowest fund weight since 2019 while the underlying spend data —
OpenAI's enterprise mix above 40%, AI-native application spend up 108%, agentic SaaS at a 46.8% CAGR
— points the other way. Fundamentals and positioning are diverging, and divergence is where the main
page's framework says to look.

The reason for caution is equally clear: the ROI reckoning is real, procurement is tightening, and
"AI compresses software pricing" is a serious argument, not a bear-case formality. The most defensible
read is that this becomes a **discrimination trade** — spend concentrating into vendors who can prove
outcomes, with a widening gap between them and everyone else, rather than a rising tide.

**The checkable test:** in the next reporting cycle, which vendors disclose AI revenue as a separate,
growing line item with retention attached, and which describe it qualitatively? That single
distinction does more work than any valuation screen in this bucket.

---

## 7. Early-stage watchlist — tracking fields

| Company / theme | Category | Narrative | Analyst sent. | HF interest | Catalyst timeline | Risk tier | Volatility | Theme alignment |
|---|---|---|---|---|---|---|---|---|
| Data layer (SNOW, MDB) | AI Software | 4 | 4 | 2 | Quarterly results; consumption metrics | Low–Medium | High | Very high — durable toll booth |
| Applied AI (PLTR) | AI Software | 5 | 4 | 3 | Q2 earnings; government contract news | Medium | Very high | High — multiple is the risk |
| Agentic workflow (CRM, NOW, HUBS) | AI Software | 4 | 3 | 2 | Results — watch for separate AI line item | Low–Medium | Medium | High — the pricing-shift test |
| Big-tech platforms (MSFT, GOOGL, AMZN) | AI Software | 4 | 5 | 4 | Quarterly capex + AI revenue commentary | Low | Medium | Medium — diluted by size |
| Observability (DDOG) | AI Software | 3 | 3 | 2 | Consumption trend in results | Medium | High | Medium |
| "Software is under-owned" thesis | AI Software | 3 | 2 | 1 | Q3 13F filings (mid-Nov 2026) | Medium | High | Very high if it re-rates |
| Private model providers (secondaries) | AI Software | 5 | n/a | n/a | Undated — funding rounds | High | n/a — opaque | Very high, inaccessible |
| Thin-revenue AI application pure-plays | AI Software | 2 | 2 | 1 | Undated | High | Very high | Low — ROI reckoning target |

---

## 8. Sentiment tables

Public investor, analyst and institutional sentiment, split three ways. HTML versions of all nine
tables across the three agents are in `tables.html`, ready to paste into a page.

### Speculative Picks (Investor Chatter)

| Company | Sector | Why Investors Talk About It | Common Narrative | Risk Level |
|---|---|---|---|---|
| Palantir (PLTR) | Applied AI | Government and enterprise AI contracts against a demanding multiple | "Investors often say it is the only company actually deploying AI into workflows at scale; sceptics reply that ~65x forward earnings prices several years of that in advance" | High |
| Private model providers via secondaries | Model providers | OpenAI and Anthropic are unlisted, so exposure is sought through feeder structures | "Investors often note the purest exposure to this bucket cannot be bought on an exchange, and that access vehicles carry heavy fee layers" | High |
| Thin-revenue AI application pure-plays | AI Software | Narrative exposure without disclosed AI revenue | "Hedge-fund commentary describes shorting firms whose AI exposure is more promotional than economic" | High |
| The "software is under-owned" thesis | AI Software | Positioning and fundamentals pointing in opposite directions | "Investors making this argument note software fell to ~6% of hedge-fund portfolios, the lowest since 2019, while AI-native application spend rose 108% year over year" | Medium |

### Long-Hold Picks (Analyst Narratives)

| Company | Sector | Analyst Commentary | Strengths | Risk Level |
|---|---|---|---|---|
| Snowflake (SNOW) | Data layer | "Analysts commonly highlight the expanded $6bn five-year AWS agreement focused on generative and agentic AI infrastructure" | Consumption model scales with usage; data gravity is hard to displace | Medium |
| MongoDB (MDB) | Data layer | "Analysts group it with Palantir and Snowflake as a leading AI software name, tied to modern AI application development" | Developer adoption; positioned where AI applications are actually built | Medium |
| ServiceNow (NOW) | Agentic workflow | "Analysts cite it among vendors proving AI agents can expand the addressable market rather than shrink it" | Entrenched enterprise workflow position; pricing shifting toward work performed | Medium |
| Salesforce (CRM) | Agentic workflow | "Named alongside ServiceNow, Intercom and HubSpot as demonstrating agent monetisation" | Large installed base to upsell agents into; established enterprise relationships | Medium |
| Microsoft (MSFT) | Platform | "Wedbush expects 2026 to be the third year of a ten-year AI cycle with US big tech as the main beneficiaries of the monetisation inflection" | AI is upside on an already-profitable base rather than the entire thesis | Low |

### Solid Picks (Institutional Interest)

| Company | Sector | Institutional Behaviour | Why Funds Accumulate | Risk Level |
|---|---|---|---|---|
| Microsoft (MSFT) | Platform | "Remains among the largest and most consistently disclosed institutional positions" | Funds describe it as owning both the infrastructure and the distribution for enterprise AI | Low |
| Alphabet (GOOGL) | Platform | "Major hedge-fund positions remain concentrated in Amazon, Nvidia, Alphabet, Microsoft and Meta" | Owns models, cloud and distribution simultaneously; internal silicon reduces supplier dependence | Low |
| Amazon (AMZN) | Platform / cloud | "Consistently among the largest disclosed institutional holdings; also took the top spot in retail investors' most-upvoted picks for 2026" | AWS captures AI workloads regardless of which model provider wins | Low |
| Oracle (ORCL) | Cloud infrastructure | "Disclosed across AI-linked infrastructure exposure in Q2 2026 filings" | Contracted AI cloud backlog gives funds a visible order book | Medium |

**A note on reading these three tables together.** The Solid Picks table is populated almost entirely
by platform incumbents, while the pure-play software names sit in Speculative. That is not an
accident of selection — it is the ~6% hedge-fund software weight showing up in table form. Funds are
getting their AI software exposure through companies where AI is upside on an existing profit base,
not through companies where AI *is* the business.

---

## Sources

- [Sapphire Ventures — 2026 Software x AI: software's AI inflection point](https://sapphireventures.com/blog/2026-softwares-ai-inflection-point/)
- [Vista Equity Partners — Software's transition to agentic enterprise AI](https://www.vistaequitypartners.com/insights/ai-impact-vista-portfolio-2026-mid-year-report/)
- [MarketScale — Enterprise AI hits an inflection point as companies rein in spending](https://www.marketscale.com/industries/software-and-technology/enterprise-ai-hits-an-inflection-point-as-companies-rein-in-spending-and-demand-real-results)
- [SaaS Mag — How SaaS companies are monetizing AI agents in 2026](https://www.saasmag.com/how-saas-companies-monetizing-ai-agents/)
- [Blossom Street Ventures — 40 SaaS earnings calls on AI](https://blossomstreetventures.medium.com/40-saas-earnings-calls-show-ai-will-be-the-biggest-boon-to-the-space-d51e93349500)
- [Tiger — Wedbush on US big tech as main beneficiaries of AI monetisation inflection](https://www.itiger.com/hans/news/2592914808)
- [Barchart — Palantir vs Snowflake analysis](https://www.barchart.com/story/news/2197151/palantir-vs-snowflake-only-1-ai-software-stock-looks-strong-for-the-next-decade)
- [Analytics Insight — Best AI software stocks 2026: Palantir vs Snowflake vs MongoDB](https://www.analyticsinsight.net/stocks/best-ai-software-stocks-2026-palantir-vs-snowflake-vs-mongodb-compared)
- [IBTimes AU — Top Palantir competitors 2026: Databricks, Snowflake, Microsoft Fabric](https://www.ibtimes.com.au/top-5-best-palantir-competitors-2026-led-databricks-snowflake-microsoft-fabric-data-ai-platforms-1865435)
- [Open — Hedge funds pile into AI stocks and chipmakers in Q2 2026 (Goldman Sachs data)](https://openthemagazine.com/business/hedge-funds-double-down-on-ai-stocks-as-chipmakers-become-wall-streets-biggest-bet)

---

*Market research and educational analysis, not personalised financial advice. Attributed opinions
belong to the parties named. Verify independently.*