What does Agentic AI mean for today’s businesses?
By Dr. Tim Sandle
SCIENCE EDITORDIGITAL JOURNAL
March 19, 2026
Many insurers remain reluctant to cover business mistakes made by 'agentic" AI programmes housed in data centres - Copyright AFP YASUYOSHI CHIBA
The “Agentic AI era” has arrived: This represents a shift in AI framed around autonomy, with the technology promising to execute tasks like lead research, intent scoring, and meeting preparation with minimal human intervention.
Whilst this is evidently exciting, experts warn that unsupervised AI can generate spammy outreach, compliance risks, and brand reputation issues.
Agentic AI refers to autonomous systems that go beyond generating content to actively achieving goals, making decisions, and managing multi-step workflows with limited human oversight. Unlike passive, prompt-driven AI, these systems, often powered by Large Language Models (LLMs), act as “digital employees” that plan, use tools, and adapt to feedback in real time to solve complex tasks
The Head of Growth/AI Product at LeadsNavi – Raphael Yu – has explained to Digital Journal how B2B marketers can leverage agentic AI for measurable pipeline lift without harming their brand, emphasizing human-in-the-loop governance, clear KPIs, and responsible integration.
Agentic AI in Sales: Where Automation Helps, and Where It Hurts
The marketing world is entering what some analysts call the “Agentic AI era.” A major AI model launch has explicitly positioned autonomy: AI acting on behalf of humans, as the next competitive battleground for sales and marketing.
What “Agentic AI” Means for Marketing
Agentic AI refers to models capable of autonomously executing tasks traditionally done by humans, such as:Researching prospects and accounts
Scoring leads based on intent and engagement
Drafting personalized outreach
Scheduling meetings or follow-ups
“Autonomy is exciting,” says Yu. “It allows teams to accelerate pipeline development, but unchecked use can backfire: sending irrelevant messages, violating compliance, or damaging brand trust.”
Where AI Adds Real Pipeline Value
Examples, from Yu, include:Research & Insights: AI can quickly scan websites, news, and social signals to identify buying intent.
Intent Scoring: Machine learning helps prioritize leads most likely to convert.
Meeting Prep & Briefings: AI drafts concise summaries, saving sales teams hours per week.
These uses translate directly to measurable pipeline lift without compromising brand integrity.
Where Agentic AI Can Hurt
Autonomy without oversight carries risks: Spammy Outreach: Automated messaging can create negative customer experiences.
Compliance & Privacy: Unsupervised AI could violate GDPR, CCPA, or internal policies.
Brand Reputation: Poorly crafted messaging can erode trust in high-value accounts.
“Agentic AI must operate with human-in-the-loop checkpoints,” Yu emphasises. “Every automated action should have governance, clear oversight, and measurable KPIs to ensure value without harm.”
Best Practices: Agentic Outbound Without Brand DamageImplement human review for all high-impact communications
Track pipeline lift separately from raw activity to measure real impact
Use AI for research, scoring, and prep, not unsupervised outbound
Enforce compliance rules and privacy safeguards in every workflow
By adopting these practices, sales and marketing teams can maximize productivity while protecting brand credibility.
The “Agentic AI era” promises faster, smarter lead generation, but the line between efficiency and risk is thin. Yu recommends a measured, governance-driven approach, ensuring AI accelerates pipeline while preserving compliance and customer trust. The real opportunity lies in autonomous assistance, not reckless automation.
Autonomy can accelerate pipeline development by handling tasks like research, intent scoring, and meeting preparation. These are measurable, high-value contributions that save teams time and increase efficiency. However, if agentic AI is used for unsupervised outbound messaging, it can create spam, compliance violations, and ultimately damage the brand.
There is great importance in human-in-the-loop checkpoints. Every automated action should have oversight, clear governance, and KPIs tied to actual pipeline impact. By separating where AI adds real value from where it can harm, marketers can safely adopt agentic AI without risking their reputation.
Yu recommends using agentic AI to assist, not replace, human judgment. For example: AI can analyse accounts, prioritise high-intent leads, and summarise research for sales teams, while humans approve messaging and maintain customer relationships.
Snowflake takes aim at AI’s follow-through problem
By Jennifer Kervin
DIGITAL JOURNAL
March 18, 2026
Photo by Visual Tag Mx
Most business leaders have a getting-stuff-done problem.
Everywhere they look, there’s data. There’s no shortage of dashboards, reports, or AI-generated insights. But someone still has to figure out what matters, pull the right data, build the output, and chase people for input along the way.
That work slows everything down, which is a no-no when you’re looking to scale.
Snowflake’s new Project SnowWork, now in research preview, is built with this battle in mind as enterprise AI that’s starting to move beyond analysis and into execution.
“We are entering the era of the agentic enterprise, ushering in a fundamentally new way to work,” says Sridhar Ramaswamy, CEO of Snowflake. “Project SnowWork looks to put secure, data-grounded AI agents on every desktop, so business leaders and operators can move from question to action instantly.”
For the past few years, across enterprise AI, the trend was for tools to focus on helping companies analyze data faster. The result has been more insight, but not always more progress.
SnowWork is designed to sit in that last mile, acting as a “proactive AI collaborator.”
“It’s about unlocking new levels of productivity and efficiency by embedding intelligence directly into the operating fabric of the enterprise,” Ramaswamy adds.
Why the ‘last mile’ matters more than the model
Many companies have already invested heavily in AI tools and data platforms.
But those investments often stall when it comes to everyday use. Employees still file requests with data teams, reports still take days, and decisions still move slower than they should.
There’s a gap between the optimism of news headlines proclaiming AI as a productivity saviour, and full buy-in for core operations.
“Enterprises have invested heavily in data platforms and AI, yet the last mile of translating governed data into everyday business outcomes remains largely manual,” says Sanjeev Mohan, principal at SanjMo.
In Project SnowWork, Snowflake is moving from “a system of insight to a system of action, which is where measurable business value is ultimately realized,” Mohan added.
In a new blog post, Ramaswamy argues that the real opportunity for enterprise AI is expanding access beyond technical teams, putting data and decision-making directly into the hands of business users across the organization.
Sridhar Ramaswamy, CEO of Snowflake (Photo courtesy of Snowflake)
“As adoption grows, a problem is emerging,” he writes. “These agents operate without shared context, governance, or coordination, making them fragmented and difficult to trust.”
SnowWork tries to address that by combining planning, analysis, and execution into a single system. It can query data, run analysis, generate outputs, and suggest next steps in one interaction.
For a sales operations team, that could mean building reports and presentations in minutes instead of days. For a marketing leader, it could mean identifying campaign gaps and generating recommended actions without waiting on another team.
It’s long past time we admit that the constant paper shuffle between teams, all in the name of completing what would otherwise be a simple project, is deeply silly.
The value in such an AI collaboration tool is reducing the coordination required to get work done.
What this changes for business leaders
In most organizations, getting from question to action still involves multiple handoffs.
Tools like SnowWork take out the ‘now what.’
First, they change who can act on data.
Tools like SnowWork are designed for non-technical users, reducing dependency on centralized data teams for everyday questions, a bottleneck that many organizations still struggle with.
That has implications for team structure. If more employees can generate and act on insights directly, the role of centralized data teams may shift toward governance, oversight, and complex problem-solving.
Second, it compresses decision cycles.
When reporting, analysis, and execution happen in one flow, timelines shrink. Decisions that once took days can happen in hours, or minutes in some cases.
That sounds incremental, but at scale it changes how organizations operate. Faster feedback loops often lead to more experimentation, quicker adjustments, and less reliance on static planning.
Third, it raises new governance questions.
SnowWork is built on “governed enterprise data,” with role-based access controls and auditability built in.
“In each case, intelligence is not just producing recommendations, it is driving action, within enterprise-defined boundaries,” Ramaswamy explains.
If AI systems are going to execute tasks, not just suggest them, companies need to be clear about who can do what, with which data, and under what conditions.
That means AI strategy is now about control.
Where work starts to move
Enterprise AI is moving into a new phase.
The first phase was about access to data. The second was about generating insights. The next phase is about execution.
SnowWork positions AI as a system that helps complete work on their behalf, like a “control plane” for enterprise AI. Systems coordinate actions across data, models, and applications instead of simply returning answers.
That idea is often described as the “agentic enterprise,” where systems can plan and carry out tasks with minimal human intervention.
“Enterprises need more than models and applications,” he writes. “They need a coordinating layer, a central control plane that aligns intelligence, enterprise data, policy, and execution across the organization to drive agentic cohesion.”
In other words, companies need AI to “show, don’t tell.”
Final shotsThe real bottleneck in AI adoption is execution. Tools that close that gap will shape the next phase of productivity.
Leaders should focus more on how work flows across teams, not just adding the latest tools. That is where most of the friction still lives.
Governance will move to a frontline priority as AI systems begin to take action instead of simply offering recommendations.
Written ByJennifer Kervin
Jennifer Kervin is a Digital Journal staff writer and editor based in Toronto.
