Building a Reliable Foundation for Agentic AI in SMBs

This article is sponsored by Salesforce and was written, edited, and revealed in alignment with our Emerj sponsored content guidelines. Learn extra about our thought management and content material creation providers on our Emerj Media Services page.

Customer-facing organizations now face a widening capability hole pushed by escalating multi‑channel demand and the constraints of human-only workflows and unsynthesized knowledge.

The U.S. Bureau of Labor Statistics projects that employment of customer support representatives will decline by 5% by 2034, whilst inquiry quantity continues to climb — a sign that automation, not hiring, is predicted to fill the hole.

On the gross sales aspect, a landmark Harvard Business Review research of two,241 U.S. companies found the common firm took 42 hours to reply to a new lead, and organizations that waited 24 hours or longer had been greater than 60 instances much less prone to qualify that lead than these responding inside the first hour. Although revealed in 2011, the research stays one of the crucial broadly cited examinations of speed-to-lead as a result of it quantified how quickly qualification charges deteriorate with rising response instances.

Consumers are already responding to the consequence: Pew Research Center finds that whereas roughly half of U.S. adults now use AI chatbots frequently, solely 29% of these customers belief the knowledge the instruments give them.

Emerj just lately hosted conversations with leaders at Salesforce, Sharif Karmally, VP, SMB Product Marketing, Matt Kravitz, Head of Customer Transformation for Service Cloud, and Vanessa Tabbert, VP of Agentic Transformation and Sales Development, in a sequence on agentic readiness for SMBs and constructing the information, workflows, and guardrails for AI‑pushed development.

This article examines the core insights SMB leaders must undertake AI brokers safely, successfully, and with measurable operational affect:

• Structured knowledge basis for dependable agent habits: Give brokers entry to unified, ruled context to stop accuracy failures and allow constant reasoning over degraded, fragmented knowledge.
• High-volume use case choice for quick, low-risk adoption: Target repetitive, low-complexity workflows to take away buyer friction at scale and ship instant operational raise with out complicated builds.
• Governed agent boundaries for protected autonomous execution: Define permissions, human checkpoints, and resolution limits to make sure brokers act as reliable contributors moderately than uncontrolled automations.
• Embedded workflow integration for seamless workforce adoption: Deploy brokers straight inside current instruments to eradicate additional steps and drive excessive adoption inside present workforce workflows.

Listen to the total episodes under:

Episode 1: Agentic CRM for SMB Automation – with Sharif Karmally of Salesforce

​

Guest: Sharif Karmally, VP, SMB Product Marketing at Salesforce

Expertise: Product Marketing Strategy, Growth Strategy, Go-to-Market Strategy, AI-Native Marketing

Brief Recognition: Sharif Karmally is VP, Global SMB Marketing at Salesforce, the place he leads world product advertising and marketing, viewers advertising and marketing, discipline advertising and marketing, and digital campaigns. He beforehand held advertising and marketing management roles at Atlan and Human Interest and led development technique, buyer retention, and monetization at Asana. He holds a B.A. in Honors Business Administration from Ivey Business School at Western University.​

Episode 2:  The Future of Customer Success and Turnkey AI Agents for SMBs – Matt Kravitz of Salesforce​

Guest: Matt Kravitz, Head of Customer Transformation for Service Cloud at Salesforce

Expertise: Customer Service Transformation, AI-Driven Service Strategy, CRM Transformation, Service Operations

Brief Recognition: Matt Kravitz is a product and buyer transformation chief with expertise spanning Salesforce, Hulu, DTiQ, Oracle, and IBM. At Salesforce, he leads Customer Transformation for Agentforce Service, working with enterprises on service technique, self-service, AI adoption, and repair maturity. Previously, as Head of Technology for Viewer Experience at Hulu, he managed a $10M+ annual expertise program supporting Disney Streaming’s viewer expertise, together with AI, omnichannel, service, and workforce optimization methods. He additionally served as CIO at DTiQ, the place he led CRM, buyer operations, analytics, and software transformation initiatives as the corporate tripled in income. He holds a BA in English from Emory University and an MBA in eManagement from Georgia State University’s J. Mack Robinson College of Business.

Episode 3:  AI-Powered Revenue Operations: The Future of Sales for SMBs – with Vanessa Tabbert of Salesforce​

Guest: Vanessa Tabbert, VP of Agentic Transformation and Sales Development at Salesforce

Expertise: Agentic Transformation, Sales Development Strategy, Enterprise Sales, Revenue Growth

Brief Recognition: Vanessa Tabbert is VP of Global Sales Development and Agentic Transformation at Salesforce, the place she has spent greater than eight years progressing by gross sales growth management roles. She beforehand served as Regional Sales Manager at MemberClicks and Director of Sales at Fathom Voice, bringing expertise in complicated gross sales cycles, gross sales workforce growth, and constructing repeatable income processes.

Structured Data Foundation for Reliable Agent Behavior

Sharif Karmally opens the dialogue by reframing SMB knowledge readiness as an operational basis drawback moderately than a technical one. He argues that many AI agent failures stem much less from mannequin functionality than from the fragmented spreadsheets, stale fields, and inconsistent buyer histories they’re requested to function on. In his view, SMBs underestimate how shortly accuracy degrades when a number of methods, inboxes, and human‑maintained paperwork turn into the de facto supply of fact.

In his episode, Matt Kravitz reinforces this framing by his agent‑maturity mannequin. He explains that agent capabilities progress from answering questions, to accessing contextual enterprise knowledge, to taking motion, making dependable context more and more necessary as autonomy grows. Level 1 brokers can purpose, however solely generically. Level 2 brokers can entry CRM or Data Cloud context. Level 3 brokers can take motion. Most SMBs try Level 3 behaviors whereas nonetheless working on Level 0 knowledge, a hole that may trigger unreliable execution.

Adding the operational penalties of scale, Vanessa Tabbert explains that her SDR (Sales Development Representative) group might solely prioritize a fraction of incoming demand regardless of working with structured lead-management processes. Roughly three out of 4 inbound leads by no means reached a human consultant as a result of the amount of alternatives outpaced what groups might realistically interact. Her expertise highlights a associated problem for SMB leaders: when buyer info, interactions, and alternatives accumulate quicker than individuals can course of them, companies want methods that may floor, set up, and act on context constantly at scale.

Their mixed steerage varieties a sensible basis SMB leaders can use to organize workflows for agentic AI:

• Unify the client report earlier than introducing autonomy: Fragmented spreadsheets and inbox‑pushed processes create context rot that brokers can’t right.
• Stabilize the information layer earlier than increasing use circumstances: Agents have to be pointed at a single supply of fact, not a assortment of partial ones.
• Sequence agent maturity in line with knowledge maturity: Actionability requires entry, and entry requires construction.
• Recognize that construction and scale are interdependent: As buyer info and interactions develop, dependable methods turn into important for sustaining protection and resolution high quality.
• Assume AI will amplify no matter context exists: Agents speed up workflows; they don’t restore them.

Sharif summarizes the structural problem:

“It’s not a single use case — it’s a map of your whole enterprise. The enterprise processes, the phases a buyer goes by, that context is a lot extra than simply the information sitting in a spreadsheet. I’ve seen this firsthand working with chief knowledge officers at massive enterprises: counterintuitively, the extra knowledge and context you add with out construction, the more severe the outcomes recover from time. It’s a phenomenon referred to as context rot. A CRM is one of the best pre‑constructed infrastructure for brokers as a result of it retains all the pieces unified and updated, so brokers don’t do issues improper.”  

— Sharif Karmally, VP of SMB Product Marketing, Salesforce

The company verify that CRM turns into a minimal viable infrastructure for brokers to behave reliably, and the standard of that construction determines whether or not AI turns into an operational asset or an accelerant of current fragmentation. ​

High-Volume Use Case Selection for Fast, Low-Risk Adoption

Early agent deployments succeed once they start with work that’s already repetitive, already effectively‑understood, and already overwhelming human groups. Across the conversations, this theme emerges as a sensible sample: the only workflows, notably people who generate recurring buyer requests and eat disproportionate workforce time, are sometimes one of the best place to start.

Kravitz makes this level by specializing in how SMB demand truly behaves: most inbound quantity clusters round a small set of predictable questions that hardly ever require deep enterprise logic. Level 1 brokers could be deployed shortly, usually in roughly a week, and instantly start decreasing demand on human groups. He cautions leaders in opposition to beginning with summarization or generative replies, which solely make sense for companies with lengthy case durations. For SMBs, the quickest path to worth is addressing the handful of interactions that dominate buyer expertise.

Drawing from Vanessa’s expertise in utilized gross sales workflows, her workforce began with the portion of the funnel that they had already deprioritized — the leads that piled up quicker than SDRs might attain them. These high-volume, low-risk alternatives gave her workforce a approach to show worth shortly with out disrupting core income workflows. She factors to the place the numbers truly moved for her:

“ When we launched, we took leads we beforehand would have achieved nothing with and booked 150 conferences in the primary month alone. Once we tuned the agent based mostly on what we had been seeing, we went from reserving 150 conferences in a month to reserving 150 conferences in a single week, with the identical high quality and amount of leads. That’s once I knew we had been onto one thing.”  

— Vanessa Tabbert, VP of Agent Transformation & Sales Development, Salesforce

Her outcomes spotlight why excessive‑quantity, low‑complexity workflows are the strongest first use case: simple to mannequin and simple to measure. More importantly, they permit groups to study shortly and construct confidence in agent habits earlier than increasing into greater‑stakes motions.

Karmally provides a complementary perspective by specializing in the place operational ache is most seen; he encourages SMBs to determine the workflow that’s already breaking beneath quantity — whether or not that’s gross sales observe‑up, service backlog, or order‑standing inquiries — and to attach solely the methods required for that single movement. Early deployments needs to be deliberately small, deliberately easy, and deliberately quick. Once the primary workflow is secure, growth turns into far simpler.

Synthesizing these insights, the company define a sensible resolution rule for SMB leaders: The proper first use case is commonly the one people have already deprioritized, and clients already really feel. Whether it’s unanswered leads, order-status requests, after-hours inquiries, or repetitive service interactions, these workflows supply a low-risk surroundings for studying, measurement, and speedy iteration.

Governed Agent Boundaries for Safe Autonomous Execution

Sharif Karmally leads the governance dialogue by highlighting the second autonomy turns dangerous, when a number of people or brokers are working in opposition to the identical buyer report. He argues that SMBs can’t depend on casual norms or tribal data as soon as brokers are appearing inside actual workflows. Clear boundaries are required to stop brokers from making selections that exceed their function, to not prohibit functionality.

Tabbert reinforces that autonomy will not be a set-and-forget functionality. She argues that brokers have to be measured, coached, and refined the identical manner leaders would handle a human worker.

Sharif’s steerage facilities on defining what an agent is allowed to do, what it mustn’t ever do, and the place human judgment should re‑enter the workflow. Without these guidelines, brokers can overwrite fields they shouldn’t contact, take actions that require human approval, or create conflicting updates that disrupt downstream processes. CRM turns into important right here, functioning because the system that surfaces disagreements, reconciles competing inputs, and maintains a trusted supply of fact.

A sensible set of operational boundaries for protected autonomy emerges from Sharif’s framing:

• Allowed actions: routine updates, repetitive duties, and predictable workflow steps that carry low danger.
• Restricted actions: pricing selections, approvals, and different business-critical modifications that require specific human authorization.
• Human checkpoints: selections requiring judgment, negotiation, or exception dealing with — particularly these tied to income or compliance.
• Reconciliation guidelines: CRM mediates conflicts between human and agent inputs, making certain the ultimate state displays authorized logic moderately than whichever replace arrived final.

This construction retains autonomy protected with out slowing down execution. Agents function like junior teammates with outlined duties, whereas people retain management over the selections that carry monetary or strategic weight. Sharif is direct about the place autonomy wants a onerous cease:

“As quickly as you’ve gotten a number of individuals — or a number of brokers — you want governance. You want guardrails for what an agent can do, and permissions for what brokers or people have entry to and may edit. You want resolution processes for when there’s disagreement, and human‑in‑the‑loop checks for probably the most vital issues.

— Sharif Karmally, VP of SMB Product Marketing, Salesforce

Kravitz extends this concept by emphasizing channel technique, arguing that organizations ought to deliberately determine which interactions belong in self-service experiences, which needs to be routed to digital help, and which nonetheless require direct human involvement.

​Embedded Workflow Integration for Seamless Team Adoption

“Tools that add work fail,” in line with Vanessa’s expertise main a excessive‑quantity SDR group. All three company notice that the extra an agent appears like a part of the prevailing toolset, the quicker groups belief it and the extra worth it delivers.

Sharif presents a Slack instance in his dialog, which illustrates how this works in apply. When buyer conversations, product updates, and inside coordination already occur in one channel, embedding the agent straight into that surroundings removes friction. Their Slack‑native agent, Teddy, reads Slack threads, identifies CRM updates, and performs them robotically — retaining the system of report correct with out asking anybody to modify instruments or bear in mind an additional step.

Vanessa provides a gross sales‑execution perspective displaying that SDRs keep away from instruments that introduce new behaviors or additional clicks. Their agent succeeded as a result of it match straight into the SDR workflow and absorbed after‑hours calls, lengthy‑tail nurturing, and repetitive outreach with out altering the workforce’s routine. Rather than changing SDR workflows, the agent prolonged them into intervals and alternatives that human groups beforehand couldn’t cowl.

A sensible integration sequence emerges from these experiences:

• Embed the place work already occurs — Slack, the service console, voice, e mail, or CRM. Adoption rises when the agent seems in acquainted environments.
• Eliminate device‑switching — the agent ought to replace CRM, floor context, and execute actions with out requiring customers to maneuver between methods.
• Let the agent observe the workflow — studying conversations, monitoring circumstances, and utilizing enterprise context to anticipate actions moderately than relying solely on direct prompts.
• Automate the low‑effort steps — knowledge hygiene, standing checks, observe‑ups, and repetitive outreach that groups routinely neglect or deprioritize.
• Augment the first workspace — turning the console or communication channel into a biotic surroundings the place the agent can counsel and execute actions in actual time.

Matt describes the very best maturity degree of this mannequin, the place the service console turns into an augmented workspace that listens and acts inside the identical pane of glass:

“The third degree is whenever you by no means depart the console. It’s not simply that I can present a contextual expertise — the console itself is saying, ‘Hey, can I assist, and may I execute actions in your behalf?’ It’s eavesdropping on the work and asking, ‘Can I verify that order standing? Can I cancel that for you?’ That’s actually the maturity mannequin: shifting from a transactional console, to a contextual one, to at least one that’s augmented and may act in your behalf with out you switching instruments.”  

— Matt Kravitz, Head of Customer Transformation, Service Cloud, Salesforce

The through-line throughout the interviews is that utilization rises when brokers match naturally into current workflows moderately than requiring groups to study new ones. Agents succeed once they improve current motions — Slack conversations, SDR outreach, service console workflows — moderately than introducing parallel ones. By embedding the agent straight into the environments groups already belief, SMBs acquire scale, accuracy, and consistency with out disrupting the rhythm of each day work.