SAP CX AI Toolkit: What's Available, What's Announced, What's Still Roadmap

SAP counted 350 AI features and more than 2,400 Joule skills across its portfolio at the end of 2025 (SAP News Center, January 2026). Most of those sit in finance, procurement, HR and industry solutions. For customer experience, the useful question is narrower: which features run in production in Sales Cloud V2, Service Cloud V2, SAP Engagement Cloud, Commerce and CDP today, which ones has SAP announced with a date, and which are still keynote material?

This post answers that question as of September 2026, based on SAP’s quarterly release highlights, the Sapphire 2026 announcements and what we have implemented in customer projects. It also covers the part that decides whether any of it pays off: the rollout sequence, the data you need first and how to measure the return.

TL;DR: The safe bets today are Joule queries and summaries in Sales Cloud V2, case classification, routing and response suggestions in Service Cloud V2, and predictive segments and send time optimisation in SAP Engagement Cloud. SAP’s packaged CX assistants were announced at Sapphire 2026 with general availability planned for Q3 2026. Roll out one feature per application, capture baselines first and measure efficiency, quality and business impact in that order.

How do we sort SAP CX AI by maturity?

We use three categories:

• Live: generally available, and we have run it in a customer tenant.
• Announced: SAP has published the feature with a planned availability date, or it is in an early adopter programme.
• Vision: demonstrated or described by SAP, without a date.

Everything below reflects the status as of September 2026. SAP ships CX releases quarterly and dates move, so check the SAP Roadmap Explorer before you plan around anything in the second or third column.

Product	Live	Announced	Vision
Sales Cloud V2	Joule queries, opportunity and account summaries, email drafts, lead scoring, deal insights, custom agents via Joule Studio	Sales, Deal Qualification and Deal Closing Assistants (GA planned Q3 2026)	Deals progressing through stages with minimal human input
Service Cloud V2	Case classification, routing, response suggestions, case summaries, knowledge recommendations, sentiment	Case Management and Service Management Assistants (GA planned Q3 2026)	End-to-end resolution of routine cases without an agent
SAP Engagement Cloud	Predictive segments, send time optimisation, product recommendations, AI report builder	Content and Campaign Assistants (GA planned Q3 2026), AI-assisted campaign duplication (early adopter)	Campaigns that retarget and reallocate budget on their own
Commerce Cloud	Product recommendations, search relevance tuning	Merchandising, Shopping and Order Management Assistants (GA planned Q3 2026)
CDP	Identity resolution, audience building		Real-time next-best-action across channels

Sources for the announced items: Autonomous CX at SAP Sapphire (May 2026) and SAP Business AI release highlights Q2 2026.

What works in SAP Sales Cloud V2 today?

Sales Cloud V2 gets more of SAP’s AI investment than any other CX product, and the everyday features are solid.

Joule as an assistant. Reps ask in plain language (“Which opportunities over 50K close this quarter?”, “What is the status with this account?”) instead of building a report. Joule answers with the rep’s own data and permissions, summarises opportunities and accounts, and drafts follow-up emails from the meeting notes and CRM context. The drafts are a good first version for routine follow-ups and need more editing for complex situations. Our Joule hands-on guide goes through it screen by screen. This is the feature we recommend switching on first.

Lead scoring and deal insights. Lead scoring ranks leads by firmographic fit, engagement and historical conversion. Deal insights flag opportunities whose activity pattern suggests they are stalling. Both depend on data discipline: if half the leads have no industry and stages are used loosely, the scores are noise.

Custom agents with Joule Studio. Joule Studio is SAP’s environment for building agents that run multi-step workflows across Sales Cloud V2, S/4HANA and other systems, for example a pre-call briefing that pulls open opportunities, service tickets and order history into one page. SAP relaunched it at Sapphire 2026 as a managed platform with free design-time access until the end of 2026 (SAP News Center). Start with one contained, repetitive workflow.

What this looks like in a rep’s day: a short Joule question replaces the Monday pipeline review, the account summary replaces fifteen minutes of clicking through tabs before a call, and the follow-up draft is ready when the meeting ends. None of that is dramatic on its own. Across 30 to 50 active accounts, it adds up to hours per week.

Announced: at Sapphire 2026 SAP presented a Sales Assistant, a Deal Qualification Assistant and a Deal Closing Assistant, with general availability planned for the third quarter of 2026. Our analysis of agentic AI in sales explains what these agents are meant to do and where a human stays in the loop. Where the standard product still has gaps, our honest assessment of Sales Cloud V2’s limitations lists them.

How mature is AI in SAP Service Cloud V2?

Service is where AI pays back fastest, because the metrics already exist: every case has a handle time, a resolution rate and a cost.

Live today:

• Automatic case classification and routing. AI assigns category, priority and product, then routes the case by content, skills and workload. Accuracy depends on how consistently your historical cases were categorised, so test it on a sample before you rely on it. Routing needs a proper skill mapping during setup.
• Response suggestions and case summaries. Joule drafts a reply from the case content and similar resolved cases, and summarises long case histories for whoever picks up an escalation. Agents use these because they save time on every case.
• Knowledge article recommendations. Relevant articles appear next to the case. They are only as good as the knowledge base behind them.
• Sentiment analysis. Flags frustrated customers for prioritisation. Good enough for routing, not for fine-grained emotion analysis.

These are not slideware. At OPO Oeschger, AI-powered ticket routing is one of four AI use cases in daily operation on Sales and Service Cloud V2: incoming service cases go straight to the right team, and nobody triages by hand anymore. Our post on digital service agents in Service Cloud V2 shows how custom Joule Studio agents extend this, for example with an automated warranty check.

Announced: a Case Management Assistant and a Service Management Assistant (general availability planned for Q3 2026), plus a partnership with Parloa that brings AI agents for customer conversations into Service Cloud. End-to-end resolution of routine cases without an agent remains a vision for most B2B service teams.

What about AI in SAP Engagement Cloud?

SAP renamed SAP Emarsys to SAP Engagement Cloud in February 2026 (SAP News Center), and the marketing AI is among the most mature in the CX portfolio.

Live today: predictive segments (likelihood to buy, churn risk, lifecycle stage), send time optimisation per recipient, product recommendations in email and on site, subject line suggestions, and the AI-assisted report builder that SAP made generally available in Q4 2025 and says cuts campaign analysis time by up to 67% (SAP News Center). Predictive segments and recommendations need several months of transaction history before they are reliable. Send time optimisation shows its effect in the next campaigns, which makes it an easy feature to prove.

Announced: AI-assisted duplication of email campaigns in an early adopter programme (Q2 2026), and a Content Assistant and Campaign Assistant with general availability planned for Q3 2026. If you are deciding between editions, our comparison of the Emarsys and enterprise editions of SAP Engagement Cloud covers what each includes.

Where do Commerce and CDP stand?

In Commerce Cloud, product recommendations and search relevance tuning are established. SAP announced Merchandising, Shopping and Order Management Assistants at Sapphire 2026, with general availability planned for Q3 2026. In CDP, identity resolution and audience building are in productive use; real-time next-best-action across channels is still the long-term vision.

In what order should you roll out AI in SAP CX?

The most common mistake is switching on twelve features across four applications in one sprint. Nobody can then tell which feature helped. We roll out in three phases:

Phase 1, weeks 1 to 4: quick wins. Joule queries and summaries in Sales Cloud V2, case classification and response suggestions in Service Cloud V2, send time optimisation in SAP Engagement Cloud. Minimal configuration, and the effect is visible within weeks.

Phase 2, months 2 to 3: features that need data preparation. Lead scoring, routing with a clean skill model, product recommendations, predictive segments.

Phase 3, from month 3: agents. Custom Joule Studio agents for well-defined workflows, and the packaged CX assistants once they are available in your tenant. Agents need mapped processes, clear decision rules and an approval step.

Before phase 1, check the data each feature depends on:

Feature	What it needs
Joule queries and summaries	Your existing CRM data model; works with little preparation
Lead scoring	Consistent firmographic fields, a conversion history with defined stages
Case classification	A substantial set of historical cases with consistent categories
Response suggestions	A maintained knowledge base mapped to case categories
Predictive segments, recommendations	Several months of transaction and engagement data per contact
Agents	Documented processes, decision rules, approval workflows

Much of the data that makes CX AI useful (stock, prices, order history, credit status) lives in the ERP. That is why the integrated S/4HANA Public Cloud and Sales Cloud V2 stack matters for AI too: an agent that drafts a delivery confirmation needs reliable availability data from the ERP, not just CRM fields.

How do you measure the return on AI in SAP CX?

Most teams measure adoption (“500 users asked Joule something this month”) instead of impact. IBM’s study of SAP customers found an average return on SAP generative AI of 6.8%, projected to rise to 12.2% (IBM Institute for Business Value, 2024). Numbers like that are averages; yours depend on your data and processes, so measure them yourself in three layers:

• Efficiency, weekly (visible within days): average handle time, triage time, time to find data, time to draft a follow-up.
• Quality, monthly (needs 30 to 60 days of data): share of correctly classified cases, whether higher lead scores actually convert more often, first-contact resolution, email engagement.
• Business impact, quarterly: cost per case, revenue per rep, pipeline velocity, retention.

Four things go wrong regularly:

• No baseline. If handle time and first-contact resolution were not captured in the weeks before go-live, nobody can prove an improvement afterwards.
• Vanity metrics. Usage counts say nothing about time saved.
• Attribution confusion. New sales training and AI in the same quarter: revenue rises, and nobody knows why. Compare a group of reps with AI against a group without, over the same period.
• Measuring too late. Six months without data is six months without course correction. Early data is noisy; collect it anyway.

What does this mean for your planning?

The most mature features share three traits: they support people rather than replace them, they work on well-structured data, and they have been through more than one release cycle. The least mature ones promise full autonomy. SAP is clearly moving there, with dedicated assistants for every CX process, but a keynote demo and a productive deployment are different things, and the gap usually closes one release at a time.

Until the packaged assistants reach your tenant, we bridge the gap with governed, human-in-the-loop agents built on SAP BTP and SAP Business AI. A small example from our own marketing is an agent that watches our blog and drafts the social posts, which a person reviews before anything goes out. Start with what works, build the data foundation, and you will be ready when the autonomous capabilities arrive. If you want a second opinion on where to start, talk to us.

FAQ

How many AI features does SAP offer, and how many of them are relevant for CX?

SAP counted 350 AI features and more than 2,400 Joule skills across its whole portfolio at the end of 2025, most of them outside CX. For Sales Cloud V2, Service Cloud V2, SAP Engagement Cloud, Commerce and CDP, the list of features we see in productive use is much shorter: Joule queries and summaries, case classification and routing, response suggestions, knowledge recommendations, predictive segments, send time optimisation and product recommendations.

Which SAP CX AI feature should we switch on first?

Start where your data is cleanest and the metric already exists. For most companies that is automatic case classification and response suggestions in Service Cloud V2, because handle time and cost per case are already measured. Send time optimisation in SAP Engagement Cloud is a close second, since the effect shows up in the next campaigns.

Is SAP’s agentic AI ready for production use in CX?

Partly. Custom agents built with Joule Studio can run narrow, well-defined workflows today, with a human approving the result. SAP’s packaged CX assistants for sales, service, marketing and commerce were announced at Sapphire 2026 with general availability planned for the third quarter of 2026. Check the SAP Roadmap Explorer for the current date before you plan around them.

Does SAP CX AI need a separate licence?

Embedded AI features and basic Joule capabilities come with the SAP cloud subscription. Premium capabilities, such as SAP-delivered agents, consume AI Units that are purchased separately, and running custom agents may require further entitlements on SAP BTP. The exact scope depends on your contract, so check it before you plan a rollout.

What data quality do SAP CX AI features need?

More than most teams expect. Lead scoring needs consistent firmographic fields and a conversion history with clearly defined stages. Case classification learns from how your historical cases were categorised. Predictive segments need several months of transaction and engagement data. If the CRM data is inconsistent, fix that first.

Do these AI features work with SAP Cloud for Customer (C4C, V1)?

C4C has some AI capabilities, but SAP’s new AI, including Joule agents and the announced CX assistants, targets Sales Cloud V2 and Service Cloud V2. SAP has not announced an end-of-maintenance date for C4C as of September 2026, but its innovation goes into V2, so AI plans are a good reason to evaluate the move. Our comparison of Sales Cloud V2 and C4C covers the differences.