Salesforce Agentic Enterprise Index: Agent Deployments More Than Double Year-Over-Year

Wazzup Pilipinas!? 

An analysis of Agentforce usage among businesses consistently leveraging agents from February 2025 to April 2026 shows how different industries deploy AI agents while building trust and recognizing ROI. 

MANILA – 10 August 2026 – Salesforce has released its 2026 Agentic Enterprise Index, which analyzes aggregated AI usage data from the Agentforce platform to uncover how businesses are deploying, using, and getting value out of AI agents. 

The results this year show a tale of two rollouts, a bifurcation in the global business landscape in which an industry’s operational DNA and business needs can dictate how its digital workforce is deployed. 

High-volume, task-specific deployments: Most common among consumer-facing sectors, this approach leans into high volume and speed to tackle immediate customer needs.

Versatile, multi-step deployments: Seen in more operationally complex and heavily regulated fields like Manufacturing and Public Sector, this strategy prioritizes building agents capable of taking on a wide variety of tasks that require cross-functional business logic.

Each type of rollout deepens trust and results in ROI, but the future of agentic AI rests on the coexistence of both deployment models—delivering fast, large-scale impact while mastering increasingly complex tasks across all industries.

● Financial Services blends speed and regulation: The financial services industry deploys some of the most sophisticated agents, yet deploys them at the massive scale more typical of consumer sectors,  proving that deep complexity and high-volume automation can coexist, especially during peak periods like tax season.

● Retail ramps up complexity on demand: Agents in consumer-facing sectors may focus on more simple tasks, but expand their depth and increase the number of actions they execute during intense holiday rushes.

Dive deeper: 

Businesses, particularly consumer industries, are deploying more agents at speed 

The average number of agents activated per organization increased nearly 3X over the past year. Once provisioned, businesses start creating those agents within an average of 2 days. And that number is decreasing month over month, down 53% across the whole analysis period.

 

But are these agents taking meaningful action? To gauge this, Salesforce has developed a metric called the Agentic Work Unit (AWU). An AWU is one discrete task accomplished by an AI agent — the point at which raw intelligence is converted into real work.

As of April 2026, Agentforce agents' AWU output is increasing by a 15% CMGR (compound monthly growth rate). 

Consumer-facing industries, especially those with high volumes of customer interaction, are demonstrating the most aggressive AWU output. Yet despite that high output, retail AI agents typically remain narrowly focused, averaging one to two actions per agent most of the year, which suggests high volumes of routine, task-specific work.

 

Trend in action: Pandora, the global jewelry brand, faces dramatic surges in customer inquiries during peak shopping seasons like holidays and Valentine's Day. To maintain its high-touch, personalized experience during these high-volume moments, Pandora deploys Gemma, an AI concierge powered by Agentforce. Gemma instantly resolves routine customer inquiries (from order status and shipping tracking to jewelry care FAQs) while offering personalized gift recommendations based on customer preferences.

By connecting directly to back-end order systems and product catalogs, Gemma delivers fast, personalized support at scale. During peak traffic, Gemma handles 60% of routine support requests while driving a 10% increase in Net Promoter Score (NPS), enabling human reps to focus on high-touch, complex interactions.

Agents become more versatile, particularly during moments of high demand

Most routine work requires agents to perform 1 or 2 simple skills, such as looking up record details or answering a customer question. But agents are increasingly tackling more complicated tasks. Today, the average agent can act on 6 skills, up from 2 at the beginning of 2025.

 

Agents become even more capable during moments of high demand. During peak shopping season, the average retail agent was able to act on 9 skills, a 350% increase that suggests they are deployed to handle more complex, multi-step customer needs as holiday demand surges. 

The data also shows agents are increasingly taking on actions across different cloud domains. For example, rather than just answering customer questions, service agents are also surfacing sales records and providing personalized recommendations, expanding the versatility of what they can do for customers and businesses. 

 

This ability to act across disparate cloud systems underscores the practical necessity of a headless architecture. By decoupling the agent's logic from traditional front-end user interfaces, agents can process tasks, execute actions, and trigger workflows anywhere.

Regulated and operationally complex industries tend to consistently prioritize agent complexity

Trend in action: Siemens sells thousands of hardware and software products across 7 siloed business units and 18,000 sellers. With 2,800 unqualified inbound leads per week and no visibility into budget, authority, or timeline, sellers wasted time chasing the wrong leads while others went untouched. In an industry defined by long, technical, multi-stakeholder sales cycles, Agentforce broke qualification into a coordinated multi-agent workflow where one agent engages and nurtures the lead, a second gathers missing data, runs it through qualification rules, and routes it with full cross-division context. That orchestration let Siemens handle a complex process end-to-end, 24/7.

To gauge the complexity of agent usage across industries, we established a "Sophistication Index" by mapping all agent actions into 5 progressive tiers of cognitive complexity: 

1. Levels 1–3 (Read, Coordinate, Synthesize): Standard, low-risk capabilities like looking up records, drafting emails, and summarizing documents.  

2. Levels 4–5 (Write, Analyze, Parse): High-complexity capabilities like updating database fields or programmatically extracting nested parameters from raw user inputs.  

By evaluating how many of these complexity levels agents in each industry actively use, the Sophistication Index highlights who is leveraging agents to take on a variety of complex tasks rather than just automating high-volume tasks. 

Manufacturing, financial services, and HLS build more advanced agent networks than traditional AI front-runners (e.g, Technology and Retail). They are deploying agents across the full spectrum of work — from retrieving and summarizing data to drafting communications and updating records directly.

 

These industries are delivering more modest AWU volume, but they are growing significantly. The AWU output of Public Sector and HLS has grown 227x and 19x, respectively.

 

Financial Services is an example of a regulated, operationally complex industry deploying AI agents at scale - representing a similar share (10%) of total monthly agent AWU output. This activity is driven by seasonal consumer surges - like Tax Day - which increase demand for agentic support.

Trend in action: PenFed operates in a heavily regulated environment. As a federally chartered credit union serving military members and their families, it must navigate strict compliance, security, and verification standards across every member touchpoint. Moving beyond traditional, rule-based chatbots required robust risk controls and cross-functional legal and compliance oversight.

To deliver seamless, automated service at scale, PenFed deployed an agent called Ace. Secured behind online banking logins, Ace acts as an intelligent assistant capable of evaluating account balances, checking loan application statuses, transferring funds, and delivering grounded answers from a curated knowledge base. Another agent, Echo, extends these multi-action capabilities to the voice channel, designed to entirely replace legacy interactive voice responses (IVR) and automated teller systems. 

PQ: “Our vision has always been to use AI in a trusted, practical, and meaningful way to serve our members seamlessly across every channel. Bringing that vision to life meant going beyond simple chatbots to build sophisticated agent experiences with trust baked in. By pairing robust governance with our unified platform, we’ve safely deployed multi-action agents like Ace and Echo that perform real, complex banking tasks—turning our goal of a truly connected, AI-enabled credit union into a reality." - Shree Reddy, CIO, PenFed

Deploying agents results in increased efficiency and higher sales growth

Across all industries, AI agents are shifting from simple conversation to execution. Agents are taking more actions (like triggering background workflows and business logic) compared to text generation. That action-to-output ratio is growing at a 15% Compound Monthly Growth Rate. For example, a service agent doesn't just draft a reply about a customer's order; it looks up the record, applies the business rule, and then issues the refund or rebooks the appointment. 

 

Businesses that leverage AI agents also see stronger sales growth compared to those who don’t. This trend is particularly strong in retail and consumer industries during holiday shopping.

 

“Whether you're spinning up agents to operate at massive scale or orchestrating them through deep, multistep pipelines, the bottom line is they’re shipping real value,” said Joe Inzerillo, Salesforce President of Enterprise AI and Technology. “That ROI isn't just showing up on the top line in sales numbers but in execution efficiency. We are moving from passive chatbots and predictive models to execution-driven agents that actually roll up their sleeves and drive real value.”

Trust is deepening - employees are using agents more and service agents are having more conversations with customers without compromising quality

The average employee engaged with an agent 300% more often per week across the analysis period over the course of the year, an indicator of trust. Slack agents average 67 sessions per week, up 3x in April from February. 

Customer Zero: Slackbot at Salesforce

Salesforce’s AI agent in Slack – Slackbot – is the fastest-adopted tool in Salesforce history. With 83% of the company using it regularly, it saves the average employee up to 5 hours of work a week. Salesforce employees use Slackbot directly within Slack to streamline daily tasks, from summarizing missed threads and drafting content to assembling briefs by aggregating metrics and files across various channels and apps.

Customers trust agents too. Over the past five quarters, agents handled 170 times more customer service chats than previous years, and consistently solved 7 out of 10 of them without needing human help. 

 

In fact, a recent study found that 77% of shoppers who engaged with on-site branded shopper agents felt more confident with their purchase than those who didn’t.

In customer service organizations, agents are making the biggest impact on customer satisfaction i — more than service rep productivity, average handle time, customer retention, and first-response time.

Looking Ahead

As the enterprise digital workforce matures, businesses won’t need to choose between agents that deploy quickly or tackle complex tasks. They will eliminate the trade-off entirely, scaling output quickly to meet surges in demand while maintaining the deep, cross-functional logic required to handle complex workflows. 

Methodology:

Powered by Agentforce and other Salesforce products, Salesforce analyzed and aggregated usage data of a cohort of businesses to uncover the true story of agents in the workforce. Looking at trends from February 2025 to April 2026, the Salesforce Agentic Enterprise Index analyzes the activity and engagement of real businesses leveraging the power of AI agents to drive ROI. To qualify for inclusion in the dataset, businesses needed to have activated agents in production every month across the analysis period. These results are not indicative of Salesforce performance. 

The Dawn of Unbroken Worlds: Alibaba Unveils Wan3.0 and Shatters the Boundaries of AI Video Generation

Wazzup Pilipinas!? 

The canvas of generative artificial intelligence has long been bound by a frustrating constraint: the fleeting glance. Mainstream video generation models have traditionally been trapped in tight loops, forcing creators to stitch together chaotic 3-to-15-second fragments just to tell a rudimentary story. Even Alibaba's preceding model, Wan2.7-Video, maxed out at this standard 15-second ceiling. 

The era of choppy cuts and fractured narratives is officially over. 

Alibaba has officially rolled out the public beta version of Wan3.0, an advanced video generation powerhouse that doubles the duration limit to an unprecedented 30 seconds per continuous clip. Designed to streamline professional workflows by merging high-fidelity creation, visual consistency, and a massive array of multimodal inputs into a single system, Wan3.0 is poised to redefine how stories are told and data is visualized.  

Breaking the Time Barrier: 30 Seconds of Unbroken Cinematic Freedom

For creators and filmmakers, the leap to a native 30-second duration is nothing short of revolutionary. By supporting longer continuous sequences, Wan3.0 allows directors and digital artists to execute complex, sweeping camera movements and unbroken tracking shots that previously required tedious post-production stitching. 

To make this extended timeline intuitive, the model introduces:

An Intelligent Duration Feature: Automatically recommends the optimal video length based on the nuances of the user's prompt. 

Advanced Video Extension Tools: Seamlessly expands narrative timelines, allowing creators to effortlessly build deeper, more immersive sequences. 

Total Multimodal Synergy: From Static Data to Dynamic Cinema

What truly separates Wan3.0 from its predecessors and industry competitors is its ravenous appetite for diverse data formats. While typical models rely heavily on simple text prompts, Wan3.0 can process text, images, video, and audio inputs simultaneously. 

Its capabilities stretch even further into the professional workspace, offering native support for web pages, PDFs, and PowerPoint presentations. This breakthrough capability completely transforms static, text-heavy corporate data, academic research, or business pitches directly into fluid, dynamic video content. 

Conquering Chaos: Precision and Realism

AI-generated video has historically struggled with visual drifting, warping, and the uncanny valley. Wan3.0 directly combats these flaws through high-precision visual continuity.  

Human Realism: Renders hyper-realistic human faces complete with synchronized, nuanced micro-expressions.  

Multilingual Audio: Produces natural, lifelike multilingual voice outputs.  

Technical Stability: Accurately presents complex, stable software user interfaces and intricate motion graphics without jitter or distortion.  

Furthermore, the model maintains strict fidelity to reference inputs. Instead of generating loose approximations, Wan3.0 meticulously replicates characters, props, spatial layouts, audio styles, and specific brand or product details. By anchoring these elements while maintaining stable layouts and audio, it transforms standard algorithmic output into deeply compelling, emotionally resonant drama.  

Beyond Entertainment: Applications Across Industries

Since the inception of Alibaba’s Wan series in July 2023, the underlying architecture has undergone continuous, rapid evolution to become more realistic, controllable, and creator-friendly. Wan3.0’s versatility now opens doors across multiple sectors:  

Filmmaking & Social Media: Streamlines production for short dramas, independent films, and high-impact social media content.  

Corporate & Education: Empowers businesses and educators to effortlessly convert static reports and text decks into dynamic marketing and teaching modules.  

Technology & Robotics: Serves as a vital simulation engine for tech developers, generating hyper-realistic video environments to train autonomous self-driving cars and advanced robotics systems. 

Step Into the Future

The public beta testing phase for Wan3.0 is officially open. Creators, developers, and enterprises looking to experience the future of generative media can apply for model testing directly through Alibaba Cloud’s AI development platform, Model Studio, and the AI-native cloud platform, Qwen Cloud.  

The boundaries of digital creation have expanded. The only remaining limit is imagination.  

When the Waters Rise: An Urgent Reckoning on Floods, Corruption, and Accountability

Wazzup Pilipinas!? 

With less than two years left in his term, President Ferdinand Marcos Jr. faces a damning and familiar reality: as the rains arrive, the water rises, and ordinary Filipino families are left to suffer the devastating consequences. Homes flood, livelihoods vanish, children are displaced, medicines are ruined, and workers lose vital income as vulnerable communities fight simply to survive.

The Human Cost of Systemic Failure

The crisis is more than just rising water—it is a profound failure of government systems meant to protect the public. A striking example of this breakdown involves a sick person with a disability who had to be carried upstairs because floodwaters had already breached their home. A society where a vulnerable citizen battling illness must be hauled to safety because public infrastructure failed to keep the flood out is a reality that demands collective outrage.

While no one can stop the rain, citizens can—and must—demand better:

Proper Infrastructure: Properly designed and maintained drainage systems that actually work.

Project Integrity: Assurance that billions allocated for flood-control projects are used to protect the communities they were intended to serve.

True Accountability: An end to empty warnings and a demand for real structural answers.

Following the Money, Following the Evidence

Corruption in flood control is never just about abstract figures on paper. Its true cost is measured in human suffering: water pooling inside a living room, a mother wading through a flood with her child, a farmer losing an entire harvest, a worker forfeiting a week's wages, and a family watching years of savings wash away.

To stop the cycle, the administration must take direct, uncompromising action:

Audit and Investigate: Deeply examine where public funds went, audit the projects, and track contracts, contractors, and subcontractors.

Trace the Beneficiaries: Find out who truly profited from public works that failed to hold back the tide.

Unflinching Prosecution: Follow the evidence all the way to the top. If government officials, businessmen, political allies, or contractors are implicated, the investigation must not stop simply because the names are powerful.

Beyond Resilience: Building a Safe Nation

Filipinos are exhausted by hollow advice to "stay safe" when communities were never engineered to be safe in the first place. Legendary Filipino resilience must never be twisted into an excuse for institutional failure.

President Marcos still has time to prove that accountability respects no political alliances, shields no powerful surnames, and leaves no position untouchable. Investigations must not merely be promised; they must be finished. Corruption must not just be condemned; it must be prosecuted.

The rain will fall again, and the floods will return. But whether those storms translate into human tragedies is not entirely beyond human control. By stamping out the corruption that makes natural weather patterns deadlier, the government can finally build a country where staying safe isn't just a wish, but a reality.