AIforce and system integrators: what SI/AI Force means

In autumn 2026 “SI / AI Force” means two things. AIforce is the layer Salesforce launched on 15 September 2026 to make its data and workflows usable from Claude, Slack and Lightning. The rename to SIForce is so far known only from secondary press; no Salesforce document confirms it. SI also stands for system integrator, and Salesforce has rewired those incentives: partners now earn on consumption and on agents in production. For manufacturers that means keeping the name AIforce, piloting small, and measuring partners on outcomes, cost caps and clear AI rules.

What does “SI / AI Force” mean?

Anyone asking about “SI Force” or “AI Force” in October 2026 usually means one of two things, often both at once:

Reading	What it refers to	Who it affects
AIforce (SIForce according to the CEO)	Interface layer that makes Salesforce usable from Claude, Slack and Lightning	IT lead, sales, service, data protection
SI as system integrator	The implementation partner who introduces and runs Salesforce for you	Programme manager, procurement, business units

The two are linked. Your partner is expected to deliver the new layer, and since 2026 Salesforce pays partners differently for doing so. That is why this chapter covers both readings. The basic terms and the four layers of AI around Salesforce are in chapter 1; the coding agents partners build with today are in chapter 4.

AIforce: the interface layer above Agentforce

Salesforce describes AIforce as “a live interface layer that brings the full power of Salesforce to wherever people and agents work” (announcement of 15 September 2026 (opens in a new tab)). In Salesforce’s architecture picture AIforce sits above Data 360, Customer 360 and Agentforce. Requests run on the user’s existing permissions; Salesforce says there is no new permissions model. The model provider is not supposed to retain business data.

Component	Contents	Status according to Salesforce
Claudeforce	Salesforce in Claude (MCP server with 37 sales skills), Salesforce Development plug-in for Claude Code with about 40 skills	Salesforce in Claude: beta for all customers; service and industry skills only announced
Slackforce	Slackforce Surfaces, Slackbot, Slack CRM, Slack Code	no availability dates
Agentforce Coworker	AI teammate in Lightning	available; Salesforce reports 100,000 activations in 35 days
Headless Toolkit	MCPs, APIs, plug-ins, skills and developer tools	no status given

For a manufacturer this is mainly a sales topic today. The 37 skills are sales skills; after-sales and field service teams have nothing to test yet. On cost, Salesforce itself writes: “Pricing and packaging are subject to change.” The Flex Credits rate card of 31 August 2026 (opens in a new tab) still lists the Headless Platform Interactions meter for agentic access as “TBA”. How external tools reach your org through MCP is covered in chapter 3, the cost side in chapter 7.

One distinction saves a lot of confusion in workshops. In the Claudeforce announcement of 27 August 2026 (opens in a new tab) Claude is a reasoning model for the Atlas Reasoning Engine and the default model in Agentforce Vibes, Agentforce Coworker and Slack. That is Claude in Salesforce. Salesforce in Claude is the other direction: CRM data shows up in the Claude app your staff use. For data protection these are two separate reviews. What an IT lead should do about it now is in the article AIforce: Salesforce in Claude and Slack explained.

AIforce and Headless 360

Salesforce Ben describes the naming sequence as Headless 360 at TDX in April, AIforce at Dreamforce, then SIForce. Salesforce’s own AIforce announcement does not mention Headless 360, yet the company keeps using the term: since Q2 of fiscal 2027 Headless 360 counts towards Agentforce ARR of more than $1.5 billion (opens in a new tab), Max editions are to include Headless 360 capacity (opens in a new tab) in future, and the rate card assigns the new meter to Headless 360 usage. My reading: AIforce is the interface, Headless 360 the billable API and MCP access underneath. In contracts, the name Salesforce prints on the order form is the one that counts.

Is AIforce now called SIForce?

All that is documented is one post. On 9 October 2026 Marc Benioff wrote on X: “The era of Super Intelligence is here. AIForce is officially SIForce.” I could not open the post itself, so the rename rests on secondary press alone. Salesforce Ben reported on 12 October (opens in a new tab) that there did not appear to be any official announcement from Salesforce and that it had asked Salesforce for comment. Up to 27 October 2026 I found no Salesforce press release or help article that uses SIForce.

The trigger, on the other hand, can be read in the primary source. Executive Order 14434 (opens in a new tab) of 29 September 2026 tells US executive departments and agencies to write “Super Intelligence” instead of “Artificial Intelligence” in their own documents. Existing regulations stay as they are, no enforceable rights are created, and private companies are not mentioned. Salesforce did not have to rename anything; it was a voluntary choice. That US federal agencies are Salesforce customers and now have to say “SI” is a plausible motive, but it is my assumption.

1. Apr 2026Headless 360 at TDX, according to Salesforce Ben
2. 15 Sep 2026Salesforce launches AIforce at Dreamforce
3. 29 Sep 2026Executive Order 14434 for US executive agencies
4. 9 Oct 2026Marc Benioff on X: "AIForce is officially SIForce"
5. 12 Oct 2026Salesforce Ben finds no official announcement

For your documents I recommend “AIforce (announced by the CEO as SIForce)” until Salesforce confirms the name in writing. Three names in about six months are also a good reason to structure architecture documents by capability rather than by product name.

SI as system integrator: what changes in the partner model

The second reading concerns your implementation partner. Salesforce rebuilt its partner program on 2 March 2026 (opens in a new tab):

• Two tiers, Summit and Select.
• 28 core competencies instead of 170 distinctions, including the increasingly demanded ones for Agentforce and Data 360.
• Incentives follow active consumption rather than seat provisioning.
• Agentforce capability is proven through sandbox validation, not just a knowledge test.
• AI governance and security guardrails are required.

The announcement gives no start date for the new tiers. On 15 April 2026 Salesforce added the Forward Deployed Engineer Partner Network (opens in a new tab), the FDE network for short. It is an exclusive group with Accenture and Deloitte as launch partners, plus Capgemini, Cognizant, IBM Consulting, KPMG, PwC, Slalom, TCS and smaller regional firms. Members get a direct link to Salesforce’s product teams and insight into the roadmap. Salesforce ties their incentives to agents reaching production and names industry depth as a criterion, manufacturing included. The claim that these firms drove one third of all successful Agentforce implementations is Salesforce’s own figure with no published method.

What the new incentives mean for you as a client

The logic has shifted from “licence sold, project billed by the hour” to “agent in production, consumption running”. Partly that is good for you, but there is a conflict.

On top of that, partners now build with coding agents themselves. That shortens work that used to be billed by the hour. Whether fixed price or outcome-based contracts fit better as a result is the subject of fixed price vs. time and materials with AI. What belongs in the contract when a partner codes with AI, meaning disclosure, data handling, IP, no production access and review, is in partners using AI coding: contract rules.

Worked example: two offers for a dealer agent

Assume a machine builder with Service Cloud, SAP and a dealer portal. It wants an agent that answers dealer questions on order status and delivery dates and asks two partners for offers. All numbers and circumstances are assumptions.

	Offer A	Offer B
Partner	long-standing partner	has Agentforce projects in production
Pricing	time and materials, estimated at 60 person-days	fixed price for the same scope plus a bonus if, after three months, 30 percent of dealer enquiries are answered without a service agent
AI tools	The offer says nothing about them.	The team works with Claude Code and the Salesforce plug-in.
Access	not stated	Hosted MCP access to the production org for testing

From an architect’s point of view I would proceed like this:

1. 01

   Ask partner A

   Whether and which coding agents the team uses. I find it unlikely that a partner uses no AI at all in 2026; either way it belongs openly in the contract.

2. 02

   Fixed price yes, production access no

   Accept B's fixed price, refuse the production access. Agents work in sandboxes and scratch orgs; releases go through the pipeline.

3. 03

   Sharpen the success measure

   B's rate only counts if wrong answers are sampled and checked and the cost per answered enquiry stays below an agreed limit.

4. 04

   Settle Flex Credits

   Decide who pays for Flex Credits, in sandbox and production, and agree a monthly cap with an alert.

5. 05

   Settle ownership

   Who runs the agent after go-live, who maintains prompts, tests and the SAP interface? With several partners an object ownership matrix helps.

After that the two offers can be compared, because AI use, access and consumption are on the table. Price, references and team then decide which partner fits better.

What should manufacturers settle with their partners now?

• Fix the name. Write “AIforce” in concepts until Salesforce confirms SIForce in writing, and describe capabilities rather than product names.
• Agree outcomes instead of hours where the scope is clear, with a metric you can measure yourself.
• Cap consumption. Expected actions per transaction, a credit ceiling, responsibility for sandbox tests.
• Write AI rules into the contract: disclosure of tools, no production access, human review, IP in the code.
• Check industry depth. Ask for agents in production for warranty, spare parts or dealer sales, not for certificates or the FDE network logo.

These points are part of my work in partner governance. Which tool fits which purpose is the subject of the next chapter, the decision matrix; how teams organise themselves around AI is covered in chapter 9. All chapters are listed on the masterclass overview.