Loading…
What We’re Seeing Across Salesforce Engagements Before Dreamforce 2026: Common Friction Points and Fast Wins
TL;DR
Dreamforce 2026 will extend the most consequential Salesforce release cycle in years. Agentforce, Headless 360, MCP, Slack’s AI overhaul, Agentforce Revenue Management, and Agentforce Contact Center all landed before September and add another layer.
Before Dreamforce 2026, audit before you add. Teams should strengthen Agentforce data foundations, establish MCP governance, improve forecast quality, and map CPQ migrations. Audit Agentforce Service knowledge against top case drivers for the fastest AI gains. Inventory contact center integrations ahead of Open CTI retirement, and identify Salesforce-dependent workflows that can move to Slack-native experiences.
The organizations that gain most from Dreamforce are never the fastest adopters. They are the ones who were ready with a solid foundation before the announcements.
The Window Before Dreamforce
Every September, Dreamforce resets the enterprise Salesforce conversation, and it sets the agenda for the twelve months that follow.
New capabilities get announced, roadmaps get extended, and organizations leave San Francisco with ambitions that are, more often than not, larger than what they arrived with.
That cycle was familiar for a decade, however; it wasn’t a problem. The gap between announcement and deployment gave enterprises time to adapt.
The 2025–26 Salesforce release cycle did not give organizations space to wait and see.
In just one year, Agentforce moved from concept to deployment. Agentforce Revenue Management (Previously Revenue Cloud Advanced) redefined quote-to-cash, Headless 360 reduced reliance on Salesforce sessions, and Slack evolved into an MCP-powered interface layer. The platform changed faster than most enterprises could operationalize it.
Dreamforce 2026 is expected to extend each of these shifts through deeper orchestration, broader agent capabilities, and new operating models.
The problem is that most enterprises have not fully resolved what the current cycle already demanded of them.
What follows in our blog post is an honest account of where that friction is most consistently concentrated and what to do before Dreamforce 2026.
Let’s dive in, shall we?
6 Patterns We’re Seeing Across Salesforce Deployments
We consistently observe a gap between Salesforce deployment and Salesforce performance. Licensing is in place, features are ready to use, and implementations are technically complete. And yet the outcomes that were scoped, forecast accuracy, deflection rates, agent efficiency, and revenue velocity, are not materializing at the expected level.
IBM’s State of Salesforce 2025–26 report put plainly:
- Only 33% of Salesforce AI initiatives{i] are meeting their ROI targets.
- The primary barrier, cited by 53% of respondents[ii], is data quality and availability.
A separate IBM study of 2,000 CEOs found that just 16% of enterprise AI deployments[iii] have scaled beyond their initial rollout. These are not organizations that haven’t started.
The cause is the infrastructure beneath it: data quality and coverage, governance structures, adoption consistency, knowledge management, and architecture decisions that were deferred at go-live and have not been revisited since.
As Agentforce, Einstein, and Agentforce Service AI become embedded in core workflows, those weaknesses become increasingly visible. AI does not compensate for poor foundations, it exposes them.
1. Agentic AI is Creating a Pilot‑to‑Production Gap
The Opportunity
Agentforce represents the most significant shift in how enterprises interact with Salesforce since the platform’s inception because it changes who, or what, can act on information without a human in the loop. Organizations investing in Agentforce are restructuring how operational decisions are made across service, sales, and revenue workflows. That is why executive conviction is high.
The Real Friction
Agentforce has generated more sustained organizational attention across our client base than any Salesforce release we can recall.
The pattern of that attention has been remarkably consistent: enthusiastic pilots, executive sponsorship, genuine conviction that this is the right direction, and then a plateau, somewhere between sandbox success and production scale, that most organizations are still sitting on six to twelve months after their initial deployment.
The plateau has three causes that appear in combination in nearly every engagement.
The first is data readiness:
Agents reason over records, and the gap between what those records contain in a controlled test environment and what they contain across the full production dataset includes:
- Inconsistent field completion
- Duplicate records that were never resolved
- Historical data that was migrated but never cleaned
Additionally, pilot teams often don’t define the minimum viable data quality needed for an agent to perform. Without those thresholds, it’s hard to move confidently into production.
The second is observability:
Organizations that have not built structured testing, versioning, and response monitoring into their Agentforce architecture before going live find themselves unable to diagnose why production performance diverges from sandbox performance.
The third is customization:
Many organizations expect out-of-the-box agents to solve highly specific business problems. In practice, the strongest outcomes come from tailoring Agentforce with custom prompts, workflows, actions, and domain-specific logic that align with actual operational requirements.
What We See
For instance, an F&P leader cut escalations by 15% and boosted efficiency by 20% through custom Agentforce solutions and Prompt Builder to predict escalation risks, surface similar cases, automate sentiment analysis, and generate contextual support responses.
A leading financial services organization achieved 3X faster complaint insights by deploying an Agentforce-powered AI agent that automated complaint analysis, trend detection, and case summarization across digital channels.
What to Do Before Dreamforce 2026
- Validate agent performance in real-world conditions by testing with messy data, edge cases, and production integrations, not just sandbox environments.
- Establish strong governance and human oversight with defined escalation paths, continuous monitoring, and clear operational guardrails to build user trust.
- Measure success using business outcomes such as resolution time, customer satisfaction (CSAT), productivity improvements, cost-to-serve reduction, and workflow throughput.
- Strengthen your enterprise foundation by ensuring unified data, resilient integrations, sufficient API capacity, and an architecture that supports multiple AI agents at scale.
- Assign clear ownership for AI operations, including prompt management, model updates, exception handling, performance monitoring, and continuous optimization.
2. Headless 360 & MCP: The Architecture Shift That Rewards Preparation and Punishes Debt
The Opportunity
At TrailblazerDX in April 2026, Parker Harris asked the audience why any employee should ever need to log into Salesforce again. Headless 360 is the platform’s answer: every capability is accessible via APIs, MCP tools, and CLI commands, with the browser and UI session no longer required for the interactions that agents can handle directly.
The key shift is that APIs and MCP are not competing approaches; they are complementary layers. APIs continue to expose the underlying business capabilities and data services, while MCP provides a standardized interface that allows AI agents to discover, understand, and invoke those capabilities without custom integrations.
Since launch, Headless 360 has processed 4.5 million MCP calls[iv] and nearly a trillion API calls. Anthropic’s own Agentforce Sales usage increased fivefold as its employees began accessing CRM data through Claude and Slack rather than a browser session.
The Real Friction
Headless 360 and MCP are exposing something that many Salesforce environments have quietly relied on for years: human intervention.
Salesforce users routinely compensate for process gaps without realizing it. They know which exception to ignore, which record to check next, and which approval path to follow when the documented workflow no longer reflects reality.
Agents cannot rely on that institutional knowledge.
For MCP: The challenge is not connectivity but orchestration and governance. Mapping Salesforce REST endpoints directly to MCP tools often leads to failures. External AI clients do not grasp the dependencies of business processes (e.g., that you must query a Contact, grab an ID, check a Pricebook, and then update an Opportunity). The agent can perform individual actions but might not know the necessary sequence to achieve a business outcome.
Additionally, many organizations postpone governance decisions until agent adoption becomes more established. By that time, issues related to data access, persona-based permissions, auditability, escalation controls, and tool exposure are already woven into production workflows.
What We See
In a Headless 360 proof of concept, we connected two Salesforce orgs using MCP-enabled tools. When a customer issue surfaced in one environment, the agent automatically surfaced a similar issue that had already been resolved in another, along with the associated context and resolution path.
The technology worked exactly as expected. The more interesting takeaway was organizational. Valuable knowledge already existed inside the business. It was simply inaccessible to the teams that needed it.
What to Do Before Dreamforce 2026
- Audit your highest-priority Salesforce data objects for API readiness, including field completion rates, validation rule coverage, and record structures that currently rely on human correction during data entry.
- Design your MCP governance architecture by defining data source access by persona, audit trail requirements, rate limits, and escalation protocols before agents begin operating across production workflows.
- Map workflows that depend on the Salesforce UI for data correction rather than simple navigation. These represent your highest-risk Headless 360 failure points.
- Define your agent trust model by specifying what agents can read, write, and execute autonomously versus what requires human approval.
- Prioritize two or three high-impact workflows where Headless 360 can deliver immediate business value, then build the required data quality and governance foundation before scaling further.
3. Agentforce Sales and Revenue Management: The Forecast Integrity Problem
The Opportunity
Agentforce Sales and Revenue Management represents Salesforce’s most ambitious bet on AI-driven commercial operations, moving from descriptive pipeline reporting to predictive and prescriptive guidance that shapes how deals are pursued, priced, and closed. Einstein opportunity scoring, AI-driven coaching, and predictive forecasting are capabilities most enterprise licenses already include. If the data is reliable, the AI improves the commercial decisions that compound across every quarter.
The Real Friction
Many revenue organizations are attempting to build AI-driven forecasting on top of signals they do not fully trust.
Forecast accuracy has always depended on pipeline discipline, activity capture, and stage consistency. Agentforce amplifies that reality because predictive systems can only reason over the signals they receive.
At the same time, many organizations continue to operate within legacy CPQ environments where years of pricing logic, custom scripts, approval rules, and business exceptions have accumulated with limited documentation. The challenge is no longer simply migration. It is recovering and modernizing institutional knowledge that has become embedded inside the platform itself.
As Salesforce’s revenue roadmap shifts toward Agentforce Revenue Management, that complexity becomes increasingly difficult to defer.
What We See
For a customer, trend monitoring time was reduced by 80–90% after integrating Agentforce Sales (Previously Salesforce Sales Cloud) with ChatGPT-powered analysis, because the underlying data was finally clean and structured enough for the AI to reason over reliably.
A high-growth API management provider transformed legacy CPQ with Agentforce Revenue Management, achieving a 35% increase in revenue efficiency, 45–60% reduction in quote-to-cash time, 25% fewer quote errors, and 40% faster deal approvals through automated workflows and unified revenue lifecycle orchestration.
What to Do Before Dreamforce 2026
- Conduct a forecast signal audit before implementing AI by identifying gaps in activity capture, inconsistent stage criteria, and forecast categories that are updated in batches instead of continuously.
- Address behavioral governance alongside technology by separating data quality challenges from platform limitations and assigning ownership of sales data discipline to sales leadership.
- Evaluate your current CPQ environment by reviewing customizations, API dependencies, technical debt, and performance constraints that could impact modernization.
- Redesign your revenue model to support subscriptions, usage-based pricing, and multi-entity operations using Agentforce Revenue Management’s native pricing capabilities.
- Normalize your product catalog by cleaning SKUs, standardizing product attributes, and creating logical product bundles that support scalable pricing.
- Modernize your pricing strategy by replacing brittle Quote Calculator Plugin (QCP) logic with declarative, scalable pricing rules that are easier to maintain and adapt.
4. Slack: The Most Underutilized Layer in Your Stack
The Opportunity
In March 2026, Salesforce announced more than 30 new AI capabilities for Slackbot[v] in the platform’s most significant overhaul since the $27.7 billion acquisition in 2021. The announcement received coverage as a product update. It should have received coverage as an architectural shift, because what changed is the role that Slack plays in the enterprise AI stack.
Slackbot is now a fully functional MCP client. It can orchestrate work across Agentforce agents, more than 2,600 Slack Marketplace applications, and more than 6,000 AppExchange applications from a single conversation thread, without the employee, the approver, or the administrator opening a Salesforce application at any point in the process.
The Real Friction
Most organizations still treat Slack as a collaboration tool rather than an execution layer. As a result, employees initiate work in Slack but complete it elsewhere, creating fragmented workflows, repeated context switching, and operational overhead that compounds over time.
More importantly, Slack readiness is inseparable from Agentforce readiness. The value of AI-powered workflows in Slack depends entirely on the underlying agent architecture, data quality, and governance model. In practice, these are not separate initiatives; they are the same transformation effort.
What We See
We see what is possible when that architecture is in place. A HCM leader came to us with support and product teams fragmented across Agentforce Sales and Slack, toggling between platforms, losing context, and manually logging into CRM for every case update. After integrating Slack–Agentforce Sales, platform switching fell by 65%, and Opportunity data accuracy improved by 70%, because sales teams could collaborate, receive updates, and act on records from a single workflow instead of moving between disconnected systems.
In another case, a software leader got the value of Salesforce Agentforce and Slack integration extended beyond automation. By bringing AI assistance, CRM context, and collaboration into the same workflow, the customer achieved 30–40% faster case resolution and a 50% increase in queries resolved within Slack.
What to Do Before Dreamforce 2026
- Inventory the three to five workflows by inventorying the three to five processes where users open Salesforce to complete a read, update, or routing decision that originated in Slack. These are your best candidates for Slack-native agents.
- Audit your Agentforce readiness by validating that these workflows have a strong data, integration, and governance foundation before introducing Slack interactions. Address any gaps before moving to production.
- Establish ownership for reusable AI skills by defining who is responsible for designing, governing, maintaining, and continuously improving the workflow definitions that agents will execute across Slack.
5. The Next Agentforce Service Battle Will Be Fought in Knowledge
The Opportunity
Agentforce Service is no longer a case management platform with AI layered on top. With Agentforce Service Agent handling autonomous digital resolution, Agentforce AI surfacing predictive recommendations mid-interaction, and Salesforce Voice bringing real-time guidance into live calls, the platform now covers the full service spectrum, self-serve, assisted, and voice, under a single AI layer. For organizations already licensed, the capability to reduce cost-to-serve and improve resolution consistency is not on the roadmap. It is available now.
The Real Friction
Most service organizations are attempting to scale AI on top of operational foundations built for human agents. Knowledge bases were written for readability rather than retrieval. Routing logic exists across multiple systems. Escalation decisions often depend on tribal knowledge rather than documented processes.
As Agentforce Service takes on a larger share of customer interactions, these weaknesses become increasingly visible. AI can only resolve issues as effectively as the knowledge, workflows, and governance structures supporting it.
Consumption-based pricing adds another layer of pressure. Inefficient workflows, poorly structured knowledge, and unnecessary agent loops no longer impact only service quality, they directly impact operating costs.
What We See
A client with Service Voice configured for Amazon Connect outage management found AI-assisted routing and escalation flows producing inconsistent outcomes during high-severity incidents. The routing logic was sound. The knowledge articles powering case classification had not been updated to reflect the current outage taxonomy. A targeted knowledge audit against the highest-severity incident types resolved the inconsistency within weeks, with no configuration changes required.
What to Do Before Dreamforce 2026
- Run a knowledge coverage audit against your top 20 case types by volume, validate whether accurate, current articles exist for each, and whether they are structured for AI retrieval, not just human readability.
- Assign ownership of knowledge operations as a distinct function, separate from case management, with accountability for coverage, currency, and tagging consistency.
- Define deflection rate targets by case type, not as a single aggregate metric.
- Establish a knowledge review cadence tied to product and policy changes.
6. Contact Centers: The Architecture Decision You Can No Longer Defer
The Opportunity
When Salesforce launched Agentforce Contact Center in March 2026, it turned the prevailing enterprise contact center architecture into a defined transition project with a deadline attached. Open CTI enters maintenance mode with a retirement date of February 28, 2028. Enterprise contact center migrations routinely take 12 to 18 months. For organizations running complex, multi-region telephony environments, the planning window for a controlled transition has already opened.
The Real Friction
For years, contact center modernization was something organizations could postpone.
Open CTI retirement changes that equation. The challenge is not deciding between Agentforce Contact Center, BYOT, or a hybrid architecture. The challenge is understanding the environment those options need to support.
In many organizations, critical routing logic exists with limited documentation. Telephony contracts have not been reviewed in years. Compliance requirements vary across regions. Integration costs are spread across multiple budgets and rarely evaluated as a single number.
Without that inventory, architecture decisions become conversations about products rather than conversations about operating models.
What to Do Before Dreamforce 2026
- Complete a contact center environment inventory before evaluating any migration path: carrier contracts and expiry dates, custom routing logic and ownership, compliance requirements by region, and the true all-in annual cost of current integrations.
- Assign a migration decision owner with authority over commercial, technical, and operational dimensions, before the vendor evaluation begins, not after.
- Model the three path scenarios (native, BYOT, hybrid) against your specific environment, not against reference architectures; the right answer depends entirely on what you have.
- Build the migration timeline backward from February 2028, accounting for your organization’s actual change management and deployment capacity.
The Case for Acting Now
The capabilities arriving in September are being built on top of the foundations that either exist in your Salesforce environment or do not.
Agentforce’s next generation of agent types will not perform better on incomplete data than the current generation does. Expanded MCP orchestration will not be accessible to organizations that have not built the governance architecture to support it. Headless 360’s continued maturation will accelerate for organizations with clean, API-ready data structures. Slack’s deepening Agentforce integration will compound the value for organizations whose agent architecture is ready and surface the gaps for those whose is not.
The pattern across every section in this piece is consistent: the near-term wins are not in adding capability before the organization is ready to absorb it. IBM’s finding that only 33% of Salesforce AI initiatives are meeting their ROI targets is not a permanent condition. It is a description of where most organizations are right now, and it is reversible through exactly the kind of foundational work that the pre-Dreamforce window is designed for.
Don’t Wait for September to Start This Conversation! Talk to Us Now!
Grazitti Interactive is a certified Salesforce Partner with deep implementation, optimization, and strategic advisory capabilities across Agentforce, Agentforce Sales, Agentforce Service, Agentforce Revenue Management, and contact center modernization. The patterns described in this piece are drawn from real engagements. So are the outcomes.
If you are heading into Dreamforce 2026 with open questions about where your Salesforce ecosystem stands and what the window before September is worth, we would like to be part of that conversation. Drop us a line at [email protected], and we’ll get back to you.
Statistical References:
[i], [ii] IBM
[iii] IBM
[iv] The Register
[v] VentureBeat
Frequently Asked Questions
1. Why should organizations prepare before Dreamforce 2026 instead of waiting for the announcements?
Dreamforce introduces new capabilities every year, but organizations that realize the most value are those with strong data, governance, and platform foundations already in place. Preparing beforehand enables faster adoption of innovations like Agentforce, Headless 360, MCP, and AI-powered workflows without delaying business outcomes.
2. What are the biggest challenges organizations face when adopting Salesforce Agentforce?
3. How can organizations improve AI ROI in Salesforce?
4. What is Headless 360, and why is it important?
5. What is the Model Context Protocol (MCP) in Salesforce?
6. How should organizations prepare for Agentforce Revenue Management or CPQ modernization?
7. Why is Salesforce data quality so important for AI initiatives?
8. How does Slack fit into Salesforce's AI strategy?
9. How can organizations prepare for Salesforce Contact Center modernization?
10. How can a Salesforce implementation partner help before Dreamforce 2026?
What do you think?
0Like
0Love
0Wow
0Insightful
0Good Stuff
0Curious
0Dislike
0Boring
RELATED LINKS