The AI-agent boom is beginning to create measurable winners and losers across the software industry.

Goldman Sachs analysts Gabriela Borges and Callie Valenti argue that the second quarter of 2026 could mark an important turning point: AI-agent adoption is no longer just a future theme, but is starting to show up directly in company fundamentals.

Their latest assessment has turned incrementally more positive on Snowflake(SNOW.US) and Palo Alto Networks, Inc.(PANW.US), while becoming more cautious on Adobe Systems Incorporated(ADBE.US), Intuit Inc.(INTU.US) and Workday(WDAY.US). Goldman remains constructive on Microsoft Corporation(MSFT.US), Shopify, Inc. Class A(SHOP.US), Cloudflare(NET.US) and Twilio(TWLO.US), which it sees as being at the beginning of multi-quarter fundamental inflections.

The bigger message for investors is that AI is doing two things at once: creating new demand for some software platforms while weakening parts of the traditional SaaS model through budget reallocation, longer purchasing cycles and changes in how users discover products.

1. AI-agent traffic is becoming large enough to move fundamentals

One of the clearest signals is coming from Cloudflare(NET.US).

Goldman cited company data showing that AI-agent traffic now represents more than 50% of total internet traffic, up from roughly 20%-30% in 2025. That threshold arrived around 18 months earlier than Cloudflare(NET.US) had originally expected.

Platforms that are naturally integrated into AI-agent workflows appear to be benefiting first.

Cloudflare(NET.US) added nearly 2 million developers in the second quarter of 2026, compared with about 1.5 million during all of 2025. Vercel’s annual recurring revenue accelerated again during the first half of the year after an earlier inflection in late 2025, while Twilio(TWLO.US) has now posted two consecutive quarters of accelerating growth as AI-related voice use cases expand.

Shopify, Inc. Class A(SHOP.US) offers another example. Goldman said its Catalog product makes product attributes easier for AI agents to retrieve, giving long-tail merchants disproportionately greater exposure in AI-powered search. The trend started emerging in the second half of 2025 and accelerated materially in the second quarter.

What investors should watch: whether a software platform is becoming part of the infrastructure that AI agents use, rather than simply adding AI features to an existing product.

2. The shift from SEO to AI search is creating a new pressure point

The same transition is hurting companies that depend heavily on traditional search traffic.

Goldman describes the shift as a move from SEO toward AI-engine optimization, or AEO. As users increasingly receive answers directly from AI systems, companies that historically relied on search engines to acquire customers may face weaker traffic funnels.

HubSpot has already shown some pressure, including faster downgrades in service tiers and seat counts. EverCommerce has also said changing search behavior created headwinds for parts of its organic customer-acquisition channels during the first half of 2026.

Goldman is incrementally more cautious on Adobe Systems Incorporated(ADBE.US) for similar reasons. The firm sees pressure at the top of Adobe’s customer-acquisition funnel as SEO becomes less important, while image-generation tools embedded in products such as Gemini intensify competition.

The unresolved question is whether large language models can recreate the conversion efficiency that Adobe historically received from traditional search.

What investors should watch: not just AI product launches, but whether AI changes the way customers discover and enter a company’s sales funnel.

3. AI spending is often replacing existing software budgets

Another important signal comes from corporate IT budgets.

Goldman’s May 2026 CIO survey found that only 33% of enterprise AI-token spending came from new budget allocations. Roughly 66% came from reallocating existing budgets, with labor and application software among the largest funding sources.

That matters because AI spending does not automatically mean the overall software spending pool is expanding.

Enterprises are also running large numbers of AI-tool experiments, often layering new products on top of existing systems. Goldman says this is contributing to longer purchasing cycles and creating an expectation that new tools should initially be available for free.

Figma(FIG.US) Make illustrates the pattern. The product began commercializing on March 13, but third-party data cited by Goldman showed MCP connections falling roughly 25% between their May 1 peak and June 1 as customers moved from free usage toward paid credits.

Goldman believes this environment creates greater pressure for Intuit Inc.(INTU.US) and Workday(WDAY.US). Salesforce.com, inc.(CRM.US) is not immune, but the firm sees it as relatively better positioned because its enterprise licensing model increasingly bundles AI capabilities into broader contracts, giving it a clearer monetization path.

What investors should watch: where AI budgets are coming from. If AI spending is funded by cutting traditional SaaS budgets, some software companies may face pressure even while overall AI adoption keeps rising.

4. Consumption pricing is making earnings harder to predict

Cloudflare(NET.US)’s latest results also highlighted another structural change: software revenue is becoming less predictable as customers shift toward consumption-based models.

Customers using pooled spending plans accelerated their usage during the second quarter, helping drive results well above expectations. Cloudflare’s pooled revenue had already exceeded 20% of new annual contract value by the fourth quarter of 2025.

The downside of this model is greater earnings variability.

Rather than simply widening its guidance range, Goldman said Cloudflare(NET.US) is discounting the midpoint of its outlook to account for a wider distribution of possible consumption outcomes.

Snowflake(SNOW.US) faces a similar dynamic. Goldman said Cortex Code remains early in its ramp, while guidance already includes some conservatism around the product’s contribution.

Microsoft Corporation(MSFT.US) Azure has also benefited from a related pricing change. Goldman said GitHub Copilot introduced variable pricing on June 1, and some customers saw GitHub bills double year over year in the second quarter. One software company valued at about $3 billion reportedly saw its bill rise by roughly 200%.

What investors should watch: guidance may become less informative for consumption-based businesses. Usage trends, customer cohorts and pricing mechanics could increasingly matter as much as headline revenue guidance.

5. AI security could become the next major differentiation point

Cybersecurity is also beginning to divide between platforms.

Goldman attributed Check Point Software Technologies Ltd.(CHKP.US)’s weaker performance mainly to an aging product portfolio and sales-organization changes, while Fortinet, Inc.(FTNT.US) benefited from the price-performance advantage of its firewall and converged-networking products, as well as some demand being pulled forward ahead of announced price increases.

Looking further ahead, attention since July has increasingly focused on new security risks created when frontier AI models break out of traditional sandbox environments.

Goldman expects the strongest long-term positioning to remain with technologically advanced platforms, particularly Palo Alto Networks, Inc.(PANW.US) and CrowdStrike(CRWD.US).

The firm is especially positive on Palo Alto Networks, Inc.(PANW.US), citing potential upside in its firewall and Chronosphere businesses, stronger thought leadership from CEO Nikesh Arora around AI security, and targeted acquisitions including Portkey, Protect and Koi.

For CrowdStrike(CRWD.US), Goldman expects the fundamental inflection to emerge later in the year, although current valuation expectations remain elevated. Okta, meanwhile, has indicated that the agentic-identity security cycle is more likely to begin in 2027.

The investment framework is changing

The main takeaway from Goldman’s work is that the software sector can no longer be viewed simply through the question of which companies are “using AI.”

Investors increasingly need to distinguish between several very different exposures:

• Companies that provide infrastructure for AI agents may see faster usage and developer growth.
• Companies dependent on traditional search acquisition may face weaker funnels.
• Application-software vendors may lose budget share as enterprises redirect spending toward AI.
• Consumption-based pricing may create stronger upside in good quarters, but also make earnings harder to forecast.
• And cybersecurity platforms capable of addressing new AI-specific risks may gain strategic importance.

AI is therefore becoming less of a single software theme and more of a mechanism for separating business models. The next stage of the cycle may be defined not by who can add AI features fastest, but by who gains traffic, budget share and monetization power as AI agents become part of everyday enterprise workflows.