THE APEX TIMES
In a new Salesforce News story, the company says many organizations stall on ROI because only one step is automated, while reviews, handoffs, and change requests remain human-scale bottlenecks. Salesforce’s proposed fix: redesign work so agents can operate across the workflow, and organize small, empowered teams to ship solutions quickly with guardrails.
Salesforce is pushing a message that is simple, and uncomfortable for many AI adopters: building fast AI agents does not automatically make a business process faster. In a company story published Tuesday, Salesforce compares modern AI deployments to a race car that can move at incredible speed but ends up stuck behind a much slower tractor. The problem is not the engine, but the road.
Salesforce says companies often implement AI by inserting an agent into a single step of a workflow, leaving the steps before and after it paced by the organization’s existing processes. In that setup, an agent can complete a task in seconds, but the overall work pauses when it hits review cycles, handoffs between people or teams, or requests for changes. The result, Salesforce argues, is that value becomes trapped at those human-scale steps, and ROI stalls out even when the agent itself performs quickly.
The company’s story also points to a broader adoption hurdle. Salesforce claims that 74% of its employees have used or interacted with agentic coding tools such as Claude Code or Codex at least once, and that usage among nontechnical employees has risen to 66%, more than triple compared with Q1. Salesforce frames this as evidence that even employees without coding backgrounds are willing to experiment with telling machines what they need, but it also warns that empowerment without structure can create risk.
A key part of Salesforce’s argument is that agents need guardrails and governance. The company says an agent without the right protections can be confidently wrong and act on that mistake at machine speed. It also notes that IT still has responsibilities around trust, integrity, and compliance, including ownership of systems of record and data semantics, such as what a customer means or what revenue means. Salesforce says those capabilities can be exposed through APIs and “MCPs,” describing the latter as mechanisms that help systems connect and interact in a controlled way.
To operationalize its approach, Salesforce highlights what it calls Rapid Innovation Squads, or RIS. The company describes RIS as a small-team model designed to accelerate “agentic transformation,” emphasizing that transformation extends beyond tooling. Salesforce says it began developing the concept in early 2026 after asking what would happen if product, business, and engineering roles could work as one small team, instead of passing work among multiple groups separated by programs and large, prewritten product requirements documents.
Salesforce says RIS is built around three elements: an empowered business stakeholder who owns the problem and drives decisions, a small group of agentic builders (one to five people, as small as possible), and a problem that is reduced to its smallest useful form. According to Salesforce, partnering with an opinionated stakeholder gets decisions about 10 times faster, and combining that with agentic coding tools yields outcomes about 30 times faster. Salesforce frames speed here as a proxy for value delivery, arguing that faster iterations tend to shorten the path from concept to shipped results.
To illustrate how this plays out, Salesforce offers multiple internal examples. One case involves SMB sellers who were dealing with multiple tools and losing time to context switching, with buying indicates such as website visits and usage spikes scattered across systems. Salesforce says a three-person RIS squad built the Sales Indicates Slack app in days using a 24-hour loop of plan, build, swarm, review, and ship, rather than a 12-month project plan, and that the app has generated 28,000 opportunities.
Another example described by Salesforce focuses on legal and engineering support workflows. The company says solution engineers previously spent two to three hours per week logging activity and that the process was error-prone. Salesforce reports that a three-person squad shipped a Slack Skill in less than four weeks that scans an engineer’s calendar, Slack data, and Salesforce systems to pre-fill a timesheet for confirmation and approval, with adoption at 80% and 23,000 weekly activities logged automatically. On the legal side, Salesforce says its legal department routes and triages about 39,000 cases per year, previously taking roughly 20 minutes per case; it adds that duplicate routing could lead attorneys to waste time on matters already being worked. Salesforce claims a small RIS squad used deterministic rules combined with Agentforce to route cases almost instantly, save $350,000 per year, achieve a 70x return that paid for itself within a month, and improve first-try lawyer selection by nearly three times.
Salesforce says this RIS model is now set to expand across the company, with an open call it calls RISFest in December. The company describes RISFest as an invitation for every Salesforce function to submit a real business problem, with the expectation that a Rapid Innovation Squad will build the solution in days, not quarters. Salesforce characterizes the effort as the opposite of a hackathon, saying the goal is shipping working solutions.
Still, Salesforce does not provide details in the story about broader rollout constraints, such as how it measures quality beyond speed, what the specific evaluation criteria are for agent correctness, or how governance is implemented across varied data and compliance regimes. The article also does not disclose any external benchmarks comparing RIS outcomes to traditional programs for all functions, nor does it specify costs, staffing levels, or longer-term maintenance requirements for the deployed agent-enabled tools.
For the next phase, Salesforce’s emphasis suggests that watch points will include whether agentic initiatives scale without introducing new failure modes, and whether the company can redesign workflows so that every step is capable of operating at agentic scale. If RISFest results in sustained adoption across diverse teams, the company’s core claim will be tested in practice: that workflow redesign and governance can turn fast agent performance into measurable business value.
A Texas official, Gina Hinojosa, has called for a halt to additional data-center projects while safeguards are reviewed, raising new questions for artificial-intelligence infrastructure plans involving CoreWeave and Oracle.
A recent comparison argues Oracle’s cloud business has expanded to about half of sales, while Accenture’s consulting work is tied to a broad swath of the Fortune Global 500. The two companies’ balance-sheet profiles, however, are portrayed as moving in opposite directions.
As third-party shopping assistants try to influence what consumers buy online, Amazon and Walmart are testing how AI-driven recommendations could reshape retail media advertising.
A reported noncash expense reflects how the value of stock warrants linked to customers can move as AMD expands its artificial-intelligence chip business.
A new forecast discussed by market outlets puts a 27% growth pace in focus for AMD, with the next quarter expected to extend momentum tied to data-center demand.
A market report claims many Chinese technology firms expect to receive only a small fraction of their orders over the next four months, highlighting how supply constraints and cross-border restrictions continue to shape the region’s AI buildout.
The rating change, reported by Yahoo Finance, points to rising optimism around Salesforce’s earnings outlook and could influence near-term sentiment for the CRM software bellwether.
A Yahoo Finance market note highlights Palantir’s performance metrics and argues the company’s push for “AI sovereignty” is resonating, though it does not provide additional operating details in the excerpt available.
A new step in Apple’s Siri strategy would broaden app access, continuing the company’s slower, product-led approach to adding AI-like capabilities to iPhones.
Despite a Citi move that cut Adobe to Sell from Neutral, the stock was higher as investors looked toward the company’s upcoming earnings report.
