Agentic AI enters the phase of fundamental realization, accelerating divergence in the software sector

Overview

The report summarizes industry and company data points from the prior three weeks to cross-validate software company performance during the 2026 off-earnings season. The core view is that adoption of agentic AI has risen to a level sufficient to affect revenue and usage at some companies, but the benefits are uneven: platforms with architectural advantages, developer ecosystems, consumption-billing capabilities, and sustained engineering innovation are improving, while software vendors reliant on traditional search-based customer acquisition, seat-based pricing, or core application procurement face pressure.

Core views

First, agent traffic and developer activity have accelerated meaningfully, benefiting platforms such as NET, TWLO, and SHOP that can support agent calls, edge computing, voice, and product-catalog matching. Second, SaaS application procurement may slow, as customers remain in the tool-experimentation phase, deal cycles lengthen, and free trials become a baseline expectation, while vendors also need to bear inference costs. Third, new consumption and flexible credit billing models increase uncertainty around revenue guidance, but may also produce larger positive surprises; SNOW's Cortex Code and MSFT's GitHub Copilot are important watch points. Fourth, cybersecurity may see stock-specific volatility in the near term, but AI model escape and agent identity risks will expand strategic security spending, with PANW viewed as the best near-term earnings setup. Fifth, application software risks are beginning to show up in results, with ADBE, INTU, and WDAY facing greater pressure; CRM is not immune, but is relatively more defensive due to AELA and AI monetization capabilities.

Analysis framework

The research uses an industry puzzle-piece approach, aggregating company disclosures, third-party usage data, product adoption rates, customer purchasing behavior, pricing changes, and management signals from the past three weeks, then mapping them across companies based on business model, product cycle, and demand exposure. The focus of the analysis is not a single usage metric, but whether AI activity can translate into overall company revenue, consumption, customer retention, or guidance upgrades.

Methodology notes

• Industry cross-validationPuzzle-piece data-point aggregation

  Combining recent scattered company and industry signals into off-earnings-season performance judgments

  The report summarizes traffic, developer, customer adoption, procurement, billing, and management information from the prior three weeks, distinguishing between industry-wide commonalities and company-specific factors.

• Cross-company mappingEarnings transmission analysis

  Mapping one company's data points to comparable companies based on business model and product exposure

  For example, using NET's consumption-billing changes to map to SNOW, changes in search behavior to map to ADBE and SEMR, and AI security demand to map to PANW, CRWD, and OKTA.

• Factor analysisGS Factor Profile

  Using growth, financial returns, valuation multiples, and composite percentiles to compare stocks

  The growth factor is based on forward sales, EBITDA, and EPS growth; financial returns are based on ROE, ROCE, and CROCI; valuation multiples cover P/E, P/B, and enterprise-value-related metrics; the composite percentile considers growth, returns, and inverse valuation percentiles.

• M&A analysisM&A Rank

  Classifying companies into three tiers based on potential acquisition probability

  Tier 1 corresponds to a potential acquisition probability of 30% to 50%, Tier 2 corresponds to 15% to 30%, and Tier 3 corresponds to 0% to 15%; Tier 1 or 2 may incorporate M&A factors into the target price.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• SNOW

  Incrementally positive

  Strengths

  The consumption-based business model and early scaling of Cortex Code could drive revenue performance above guidance.

  Weaknesses

  Data on the consumption queue for new products remains limited, reducing forecast precision.

  Comparison

  The earnings setup is better than DDOG, because DDOG previously had a higher expectations bar and was affected by large-customer-specific factors.

  Risks

  Cortex Code adoption or consumption conversion falls short of expectations.

• PANW

  Incrementally positive

  Strengths

  Both firewalls and Chronosphere may contribute upside, and the company is directly positioning for AI risks through acquisitions such as Portkey, Protect, and Koi.

  Weaknesses

  Near-term demand and orders in the security industry may still fluctuate.

  Comparison

  Viewed as the security company most likely to deliver strong results during the off-earnings season.

  Risks

  Near-term industry vacuum, acquisition integration, and delayed realization of AI security budgets.

• NET

  Continue to overweight

  Strengths

  Agent traffic, developer growth, edge computing, and flexible credit consumption jointly drive fundamental improvement.

  Weaknesses

  As consumption-based revenue increases, quarterly guidance and the magnitude of upside surprises become harder to predict.

  Comparison

  Changes in its consumption model provide a reference for earnings read-throughs to companies such as SNOW.

  Risks

  Consumption pace slows, guidance volatility, and high-valuation expectations.

• SHOP

  Continue to overweight

  Strengths

  Catalog can match product attributes with specific agent searches, clearly benefiting long-tail merchants, while Sidekick adoption is growing rapidly.

  Weaknesses

  The company is not currently materially monetizing AI features directly.

  Comparison

  Among application vendors disclosing growth in agent adoption, SHOP has most clearly improved its overall fundamentals.

  Risks

  Traffic and transaction conversion from agent search cannot be sustained.

• TWLO

  Continue to overweight

  Strengths

  AI voice use cases are scaling, and growth has accelerated for two consecutive quarters.

  Weaknesses

  Growth still depends on customer usage and the continued expansion of new communication scenarios.

  Comparison

  Compared with traditional seat-based application software, it benefits more directly from agent calls and communication traffic growth.

  Risks

  Usage growth slows or AI voice monetization falls short of expectations.

• MSFT

  Continue to overweight

  Strengths

  GitHub Copilot variable pricing increases customer bills and may continue to drive Azure revenue.

  Weaknesses

  Variable pricing may increase customers' cost sensitivity.

  Comparison

  It has a complete monetization chain from developer tools to cloud infrastructure.

  Risks

  Customers optimize usage, bill growth slows, or cloud demand falls short of expectations.

• ADBE

  Incrementally negative

  Strengths

  It has mature creative tools, a strong brand, and a large user base.

  Weaknesses

  The shift from SEO to AEO may weaken top-of-funnel consumer acquisition, and experimentation with generative image tools also intensifies competition.

  Comparison

  Compared with platforms exposed to agent traffic or consumption billing, ADBE is more vulnerable to changes in search channels.

  Risks

  LLM channel conversion rates cannot match traditional SEO, and free or integrated generative tools erode new-user additions.

• WDAY

  Incrementally negative

  Strengths

  Its position as an enterprise system of record and customer base provide some defensiveness.

  Weaknesses

  Core products may be affected by budget reallocations, procurement slowdowns, and seat pressure.

  Comparison

  CRM, with AELA bundling and progress in AI monetization, may be better able than WDAY to withstand industry pressure.

  Risks

  Core subscription downgrades, longer deal cycles, and insufficient monetization of AI features.

• CRWD

  Near-term cautious, long-term positive

  Strengths

  It is technologically leading and is expected to benefit from long-term demand for enterprise AI security tools.

  Weaknesses

  The expectations bar ahead of earnings is high, and fundamentals are more likely to inflect later in the year.

  Comparison

  The near-term earnings setup is weaker than PANW, but it remains a high-quality AI security platform over the long term.

  Risks

  A near-term earnings vacuum or high expectations lead to share-price volatility.

• OKTA

  Near-term cautious

  Strengths

  The identity security business has long-term strategic relevance to agent identity management.

  Weaknesses

  The company has consistently indicated that agent identity is more likely to be a CY27 product cycle.

  Comparison

  The realization of AI security revenue is later than for platforms such as PANW.

  Risks

  Product cycle delays and a lack of near-term fundamental catalysts.

Key data

• Agent traffic shareOver 50%Share of total internet traffic disclosed by Cloudflare, above approximately 20% to 30% in 2025 and 18 months ahead of the original forecast.
• Cloudflare new developersNearly 2 million in 2Q26Compared with approximately 1.5 million for full-year 2025, developer growth accelerated meaningfully.
• Sources of enterprise AI budgets33% new budget, 66% budget reallocationResults from Goldman Sachs' May 2026 CIO survey; major sources of reallocation include labor and application software.
• Shopify Sidekick usageDaily active merchants up 3.6x year over year; daily sessions up 4.8xSidekick handled nearly 34 million conversations, and merchants created more than 36,000 custom apps, above 12,000 in 1Q.
• HubSpot agent product adoptionData Agent has more than 16,000 customers; Prospecting Agent has nearly 17,000 customersUp 80% and 28% quarter over quarter, respectively; Customer Agent customer count exceeds 10,000.
• Klaviyo agent adoptionComposer Agent added 95,000 users in its first monthCustomer Agent adoption increased 40% quarter over quarter.
• Figma Make connection changeJune 1 down 25% from the May 1 peakAs shown by third-party data; customers became more cautious in spending after shifting to paid credits.
• Cloudflare flexible credit revenueMore than 20% of new annual contract value in 4Q25This model widens the distribution of guidance outcomes, potentially increasing the magnitude of near-term upside surprises but reducing predictability.
• Impact of GitHub Copilot variable pricingSome customers' 2QCY bills increased 100%, and individual customers' bills increased as much as 200%Variable pricing was introduced on June 1, 2026 and may continue to benefit September-quarter Azure revenue.
• Goldman Sachs global equity coverage3,104 securitiesAs of July 1, 2026; rating distribution was 50% Buy, 34% Neutral, and 16% Sell.

Impact & implications

The investment implication is that the software sector can no longer be traded solely on AI product launches or usage; investors should focus on whether adoption can translate into overall revenue growth, accelerated consumption, and stable monetization. Companies with infrastructure positioning, consumption billing, developer stickiness, and architectural advantages are more likely to see multi-quarter fundamental inflection points; traditional application software may come under pressure from budget reallocations, seat compression, free trials, and changes in customer acquisition channels. In the near term, investors should prioritize SNOW and PANW's off-earnings-season performance, continue to own MSFT, SHOP, NET, and TWLO, and remain cautious on ADBE, INTU, WDAY, and high-expectation CRWD.

Risks

• Enterprise AI budgets mainly come from reallocations of existing budgets, which may further squeeze application software spending.
• Customers are in the stage of product experimentation and searching for product-market fit, and deal cycles may continue to lengthen.
• Free trials are becoming a baseline expectation, and software vendors need to bear inference costs, which may compress margins.
• Consumption billing and flexible credit models widen the distribution of revenue outcomes, making quarterly guidance and the magnitude of upside surprises harder to predict.
• The shift from SEO to AEO may weaken the efficiency of traditional search-based customer acquisition, affecting the top of the funnel for companies such as ADBE, HUBS, and SEMR.
• The security segment may experience a near-term demand vacuum, and companies with high expectations may see significant volatility even if the long-term thesis remains sound.
• High adoption rates for AI products do not necessarily translate into improvements in overall company revenue or profit.
• The report is based on public information and recent data points available as of the publication date, and the relevant judgments may change with company disclosures.

What to watch

• SNOW's Cortex Code consumption growth and the magnitude of upside versus guidance when it reports results in early September.
• PANW's firewalls, Chronosphere, and AI security acquisitions and their contribution to near-term results.
• Consumption pace among NET flexible credit customers, the way guidance is discounted, and the sustainability of developer growth.
• Incremental Azure revenue from GitHub Copilot variable pricing for MSFT in the September quarter.
• Whether SHOP's agent search traffic can continue to convert into long-tail merchant orders and platform revenue.
• Downgrade rates, seat counts, and deal cycles at HUBS, WDAY, and other application software companies.
• ADBE's AEO customer acquisition efficiency and the impact of generative image tool competition on new users.
• AI security and agent identity product inflection points for CRWD and OKTA from the second half of 2026 to CY27.
• Whether AI feature adoption metrics can further translate from usage into overall company revenue, retention, and profit improvement.