What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
In a January 31, 2024 VentureBeat interview, Clara Shih described AI as a “moving target” because models, research, products, risks and customer expectations were changing unusually quickly. Her answer was not to wait for stability. It was to keep a fixed business objective—use AI to improve real customer and employee workflows—while changing the technical approach whenever evidence required it.
That distinction still matters in 2026. Salesforce’s terminology has moved from EinsteinGPT and Copilot toward Agentforce, but the underlying strategic question remains: can an organization ship useful automation now, redesign its platform for AI, and preserve enough experimentation to adapt later?
What Shih meant by “AI is a moving target”
Shih was referring to several moving parts at once: model capabilities, research findings, prompting and retrieval methods, vendor competition, product expectations, and enterprise governance requirements. A model that appears best this quarter may be surpassed soon; a workflow that works in a demo may fail when permissions, stale data or unusual customer requests are introduced.
Her practical conclusion was a two-part operating rule: commit to the business problem, not to one model or permanent interface. Companies should maintain a roadmap, ship bounded use cases, monitor technical progress and revise, extend or abandon experiments as results change.
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →#1 Best Overall
The steady aim: useful AI inside existing work
The objective Shih described was practical rather than futuristic: remove repetitive work, help service and sales employees find product knowledge, and leave people more time for judgment, relationship-building and complex problem-solving. That means embedding AI in Salesforce clouds and workflows instead of treating it as a disconnected chatbot.
The distinction is important for buyers. A writing assistant can produce text without improving a process. An integrated system can retrieve authorized customer context, recommend a next action, route work, preserve approvals and record what happened.
The Gucci meeting that changed Shih’s view
Shih told VentureBeat that a pivotal moment came during a November 2021 meeting with an Italian Gucci delegation during the COVID-19 pandemic. Gucci wanted high-touch, accurate customer-service help but did not want customers to experience a rote chatbot. Salesforce chief scientist Silvio Savarese demonstrated CodeGen, and Shih said the large-language-model capabilities made the opportunity tangible.
The interview says Salesforce had worked on CodeGen since 2018, introduced it publicly a few months after that meeting, and described an open-source model reaching as many as 16 billion parameters at the time. Those are historical interview claims, not evidence that CodeGen is a current Salesforce offering.
The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Rank #2
Shih also described AI coaching for service representatives: helping them learn products so they could become stronger sales and brand representatives. The account does not provide independently verified conversion, revenue, handling-time or return-on-investment figures, so it should be treated as a pilot story rather than a quantified business case.
Why Salesforce appeared to move quickly after ChatGPT
Salesforce introduced EinsteinGPT in March 2023 and integrated generative-AI capabilities across several clouds and products. Shih said the visible launch followed approximately 15 months of underlying work. That claim illustrates a common enterprise pattern: apparent speed may reflect earlier research, infrastructure, customer pilots and workflow testing rather than a product built in only a few months.
At the time of the interview, Shih was described as Salesforce’s first head of AI, appointed in March 2023. The available source does not establish her exact Salesforce title or role in August 2026.
Shih’s three-horizon execution model
Horizon 1: Ship immediate utility
The first horizon targets bounded departmental problems in sales, service, marketing, commerce and Slack. Examples include drafting responses, summarizing records, retrieving approved knowledge and automating routine administrative work. A good first project has defined users, inputs, outputs, human review and measurable success criteria.
Windows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallCrashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteHorizon 2: Redesign the platform around AI
The larger ambition is not an assistant bolted onto Sales Cloud or Slack. It is to rethink how work is routed, how data is retrieved, which actions require approval, and how outcomes are measured. This is the difference between adding an AI feature and making a product AI-native.
Horizon 3: Keep experimenting
Research reading, hackathons, prototypes, specialized models and conversations with founders preserve optionality. This track lets a company test new approaches without allowing every new model release to disrupt production commitments.
Then and now: EinsteinGPT to Agentforce
Salesforce’s current documentation uses more agent-oriented language than the 2024 interview. Agentforce is presented as an agent-driven layer of the Salesforce Platform for workflows across sales, service, marketing, commerce, Slack and related products. See Salesforce’s Agentforce overview.
| Historical context | Current documented context |
|---|---|
| EinsteinGPT, Einstein Copilot, Copilot Studio, Prompt Builder, retrieval-augmented generation and the Einstein Trust Layer were prominent in the 2024 discussion. | Salesforce now emphasizes Agentforce, Data 360, the Einstein Trust Layer and employee or service agents. |
| The interview discussed “topics” in the then-current product vocabulary. | Salesforce says “topics” became “subagents” beginning in April 2026. |
| EinsteinGPT represented the rapid-response product phase after ChatGPT. | Salesforce says Agentforce (Default) stopped receiving new features and was unavailable in new environments beginning June 17, 2025, with migration to Agentforce Employee recommended. |
| Product availability and commercial terms were still developing. | Summer ’26 release notes said Agentforce was planned to be enabled by default for eligible organizations in August 2026, with no change to billing. |
These later product decisions can be read as consistent with the “embedded in workflows” direction, but the 2024 interview does not prove that Shih predicted every Agentforce decision.
Enterprise reality: what must work before deployment
Grounding and data quality
Agent behavior depends on authoritative CRM and operational context. Salesforce documentation points to Data 360 and organization setup for grounding. Buyers should check stale knowledge, duplicates, missing fields, conflicting records and field-level permissions before expecting reliable answers or actions.
Security is shared responsibility
Salesforce describes its Einstein Trust Layer as including grounding, masking, toxicity detection, audit trails, access-control preservation and zero-data-retention arrangements with certain third-party model providers. Those controls do not remove customer responsibility for permissions, configuration, connected systems, prompts, agent actions and business-process governance. See the Trust Layer documentation and Salesforce’s shared-responsibility guidance.
Operational limits are product requirements
Salesforce documents 60-second action timeouts, 30-second reasoning-engine timeouts and truncation when agent-action output exceeds 65,000 characters. Multi-step workflows should be tested against those limits rather than assumed to behave like an unconstrained chat session. See Agentforce considerations.
Pricing is not one simple “AI price”
Salesforce documents consumption-based, hybrid and business-metrics-based AI pricing. Actual cost can depend on edition, licenses, agent type and usage; implementation, data preparation, governance and change management are separate budget items. The usage documentation is the appropriate starting point.
Free tools Windows power users keep installed
One-click scans. No signup required.
Best Value
What the strategy gets right—and what it cannot prove
The durable insight is organizational: ship a narrow useful product, redesign the larger platform and maintain a research track. The Gucci account shows customer-led discovery, not universal reliability. The interview does not establish how much data preparation a pilot required, which failure cases occurred, whether labor was reduced, or what return on investment customers achieved.
It also does not establish that every organization should choose Salesforce. Native Agentforce is a plausible fit when CRM data, permissions and workflows already live in Salesforce. Microsoft may fit organizations centered on Microsoft 365, Teams, Azure and enterprise identity; ServiceNow may fit IT and employee workflows; standalone model APIs offer more architectural flexibility but shift integration, evaluation, security and operations to the customer.
A practical buying test
- Choose one measurable bottleneck. Define the workflow, users, inputs, outputs and baseline metric.
- Verify authoritative context. Inventory CRM, knowledge and Data 360 quality, permissions and retention requirements.
- Run a bounded pilot. Keep human review for consequential decisions and document failure cases.
- Measure the whole process. Track resolution time, quality, adoption, escalation, error rates and supervisor workload—not just generated text.
- Plan the platform path. Decide whether the pilot can eventually redesign routing, approvals and execution, rather than remain a costly add-on.
- Budget for variability. Model licenses, consumption, implementation and model changes, and preserve an exit or portability plan.
For a company already running Salesforce, the sensible commercial move is a narrow Agentforce pilot with explicit data, governance and usage budgets—not a purchase justified by a chatbot demonstration.
The durable lesson
Shih’s phrase separates a volatile technical target from a stable business aim. Models and product names will continue to change. An enterprise that can connect trusted data, govern actions, measure outcomes, redesign workflows and keep experimenting is better positioned than one that simply chases the newest model.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
