Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Scan for outdated or missing drivers - takes under a minute3Clear out junk files and repair common Windows errorsOrganizations are investing in AI and planning customer-facing pilots, but those moves alone do not create a joined-up customer experience. Results depend on whether AI can use current knowledge, relevant customer context, coordinated channel routing, and a clear path to human help.
Why can more customer-service AI still feel like worse service?
Buying or piloting an AI tool is an adoption decision. Whether a customer can get a reliable answer without repeating information, being sent in circles, or struggling to reach a person is an orchestration question. The distinction matters because a chatbot can work as designed and still fail the customer if its answer is outdated, it cannot see relevant history, or the next channel starts the interaction from scratch.
As an Amazon Associate I earn from qualifying purchases.
Survey findings also measure different things: leaders’ plans, organizations’ investments, and customers’ willingness or preferences. They should not be collapsed into a single adoption or acceptance trend.
What the surveys show—and what they do not
| Source and population | Reported finding | What it supports |
|---|---|---|
| Gartner, customer survey of 5,728 people conducted in December 2023; reported July 9, 2024 | 64% said they would prefer companies not use AI for customer service; 53% said they would consider switching if they learned a company was going to use AI for service. | AI use can raise trust and choice concerns. These answers do not establish that every customer rejects every AI use. |
| Gartner, survey of 187 customer-service leaders conducted July–August 2024; reported December 9, 2024 | 85% said they would explore or pilot customer-facing conversational GenAI in 2025. In the same leader survey, 61% reported a backlog of knowledge articles to edit, and more than one-third lacked a formal process for revising outdated articles. | Planned exploration was not proof of deployment or effectiveness. The knowledge findings show why pilots need content ownership and maintenance, not only a model or interface. |
| Gartner, customer survey of 4,879 people conducted January–February 2025; reported June 25, 2025 | 51% said they were willing to use a GenAI assistant for customer-service interactions on their behalf. | This is a different question from the 2023 preference question, so the two percentages do not establish a clean change in customer acceptance. |
| Deloitte Digital, survey of 600 leaders responsible for contact-center strategy at midsize and large B2C and B2B companies in the United States, Australia, Canada, Japan, and the United Kingdom; conducted March 2024 and reported in a May 2024 brief | 25% of surveyed organizations had implemented an omnichannel routing engine; 76% said agents were overwhelmed by systems and information. | Channel-specific routing does not necessarily connect experiences across channels. Tool proliferation and routing gaps can coexist. |
| Salesforce, 2024 State of Service summary; over 5,500 service professionals in 30 countries, surveyed December 8, 2023–January 22, 2024 | 79% of organizations had invested in AI and 81% used workflow or process automation; 83% of service decision-makers planned to increase data-integration investment over the following year. | Investment and automation can coexist with a need to connect data and workflows. The planned integration figure is intent, not confirmation that the investment happened. |
| Avaya release of March 25, 2025, summarizing a Forrester Consulting study commissioned by Avaya | 45% planned to implement more advanced capabilities such as orchestration within the next 12 months; 37% cited replacement cost for existing technologies and 35% cited security and data privacy as concerns; 76% said phased AI adoption was critical to service quality. | These are findings as summarized by Avaya from a commissioned study, not an independent endorsement by Forrester. They point to implementation constraints and a preference for staged change, not proof of outcomes. |
The surveys differ in date, population, sponsor, wording, and whether they ask about preferences, willingness, current investment, or future plans. In particular, a 2025 pilot plan does not establish what was ultimately deployed or whether service improved.
#1 Best Overall
What orchestration means in customer experience
Here, orchestration means coordinating an AI system with the information and people needed to complete a service interaction. It is not a formal definition from a single standard; it is a practical way to assess whether the whole service journey works together.
- Customer context: make relevant identity, interaction history, and issue details available where policy permits, so the customer does not have to rebuild the story at every step.
- Maintained knowledge: ground answers in content that has an owner, a review process, and a route for correcting outdated or missing information.
- Channel coordination: route work across the channels the organization supports instead of treating each channel as an isolated queue.
- People and policies: define when AI can act, when it must defer, what an agent can see, and how privacy, security, and continuity requirements apply.
- Service outcomes: monitor whether issues are resolved with reasonable effort, not only whether automation is used or a queue is handled faster.
Salesforce’s integration and workflow findings, Deloitte Digital’s routing and agent-overload findings, and Gartner’s knowledge-maintenance findings describe different parts of this coordination challenge. They do not prove that one architecture or product is right for every organization.
How to design a useful AI-to-human handoff
A handoff is part of the service, not an escape hatch to bolt on after the bot is built. Gartner’s Keith McIntosh described the expectation this way: “For example, AI-infused chatbots must communicate to the customer that they will connect them to an agent in the event that the AI cannot provide a solution. It must then seamlessly transform into an agent chat that picks up where the chatbot left off.”
- Set a clear boundary. Define the cases the AI can handle, the signals that require a person, and the point at which uncertainty or repeated failure should trigger escalation.
- Tell the customer what is happening. Explain that the AI cannot resolve the issue and make the available human route visible. Do not imply that escalation is available if it is not.
- Carry useful context forward. Give the agent the conversation and relevant customer or case details that the organization is allowed to share, rather than making the customer repeat them.
- Keep responsibility clear. Make it apparent whether the customer is still interacting with AI or has reached an agent, and ensure the agent can take ownership of the unresolved issue.
- Review failed journeys. Use unresolved cases and repeated handoffs to identify gaps in knowledge, routing, or policy; revise the underlying service rather than merely changing the bot’s wording.
A handoff that preserves context can reduce avoidable repetition, but it cannot compensate for an agent who lacks authority, an unavailable channel, or a broken process behind the conversation.
Rank #3
How to move from pilots to a connected service
Start with a bounded customer problem
Choose a common, well-understood task and define what counts as completion from the customer’s perspective. Include a safe route to a person for exceptions. A pilot that measures only how many conversations AI handles risks rewarding containment even when customers remain stuck.
Make knowledge an owned operation
Assign owners to the information AI uses, define how often it is reviewed, and give staff a way to flag errors or missing answers. Gartner’s 2024 leader survey found both an editing backlog and organizations without a formal revision process, so content readiness should be checked before expanding a pilot.
Rank #4
Connect the journey, not just the channels
Map what happens when a customer moves from self-service to chat, phone, or another supported channel. Check whether routing rules, case state, and relevant history travel with the issue. Deloitte Digital notes that existing channel-specific routing tools do not necessarily connect experiences across channels; adding another channel is not itself omnichannel orchestration.
Phase rollout around risk and readiness
Avaya’s release, summarizing the study it commissioned from Forrester Consulting, reported that 37% cited the cost of replacing existing technologies and 35% cited security and data privacy as concerns; the same summary reported that 76% considered phased AI adoption critical to service quality. Treat those as signals to assess dependencies and safeguards, not as universal benchmarks. A phased approach can expose integration or policy problems before they affect a larger share of customers.
Best Value
Give agents a role the system can support
Automation should make complex or sensitive cases easier for staff to handle, not surround them with disconnected tools. Gartner analyst Brad Fager said: “Successful teams will shift from reactive human requests to proactive customer experience orchestration. The focus of customer service will move from managing demand to value creation, with AI supporting human agents and freeing them for expanded roles,” The aspiration is useful, but teams still need to test whether the actual workflow reduces fragmented work and gives agents enough context and authority.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.How to evaluate whether orchestration is working
Compare approaches against the service journey, not just the AI feature list. Establish a baseline for the cases in scope, then examine customer and operational evidence together.
- Resolution and effort: Is the issue resolved, and how many steps or repeated explanations does the customer encounter?
- Continuity: Can an agent pick up the prior interaction and relevant history when a case changes channel or needs escalation?
- Knowledge quality: Are answers traceable to current, owned information, and can staff correct a gap without waiting for a major system release?
- Channel coordination: Do routing and case state follow the customer across supported channels, or does each queue behave independently?
- Data protection and resilience: Is only appropriate information used and transferred, with clear security, privacy, governance, and continuity controls?
- Agent workload and service quality: Does the change reduce fragmented work while preserving the human capacity needed for difficult cases?
Track efficiency alongside customer outcomes. A system that shortens an interaction but increases repeat contacts or makes human help harder to reach has not demonstrated a better experience.
Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Clear out junk files and repair common Windows errorsFree Scan →What the evidence cannot establish
The cited surveys do not establish a universal causal result that orchestration always outperforms adoption, or that orchestration by itself guarantees better customer experience. Gartner and Salesforce published summaries of their own research; Deloitte Digital reported a survey commissioned by its practice; and Avaya summarized a study commissioned from Forrester Consulting. Their findings should remain attached to their specific populations, questions, dates, and sponsors.
Nor do leadership intentions show that planned pilots became deployments, that deployed systems worked well, or that customers experienced improved service. The defensible conclusion is narrower: investment and pilot activity are visible, while the evidence also points to coordination, knowledge, context, and handoff problems that implementation must address.
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.
