> ## Documentation Index > Fetch the complete documentation index at: https://usefulai.com/llms.txt > Use this file to discover all available pages before exploring further. # Best AI Use Cases for Customer Support in 2026 > The 7 best AI use cases for support teams - grounded replies, routing, summaries, QA, and knowledge upkeep - each with integrations and exact setup steps.
Updated July 24, 2026
Support teams lose time searching for approved answers, rewriting routine replies, and reconstructing long conversations before they can solve the customer’s problem. These are the seven AI workflows that improve that work without hiding the handoff to a human.

All 7 run on whichever AI agent you already use - ClaudeClaude, ChatGPTChatGPT, GeminiGemini, or Microsoft CopilotMicrosoft Copilot - connected to the tools listed with each use case.

## Best AI Customer Support Use Cases
| # | Use case | Key integrations | Est. time saved About the estimate | | - | ----------------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | | 1 | Draft support replies | ZendeskIntercomHelp deskZendeskConfluenceKnowledge base | **3** hr/wk | | 2 | Resolve routine requests | ZendeskConfluenceKnowledge baseZendeskIntercomHelp desk | **2.5** hr/wk | | 3 | Find approved answers | ZendeskConfluenceKnowledge baseZendeskIntercomHelp desk | **2** hr/wk | | 4 | Review support quality | ZendeskIntercomHelp deskMaestroQAZendesk QASupport QA | **2** hr/wk | | 5 | Triage support tickets | ZendeskIntercomHelp deskSalesforceHubSpotCRM | **1.5** hr/wk | | 6 | Summarize conversations | ZendeskIntercomHelp deskSalesforceHubSpotCRM | **1.5** hr/wk | | 7 | Draft knowledge updates | ZendeskIntercomHelp deskZendeskConfluenceKnowledge base | **1** hr/wk |
***

Draft support replies

Best first workflow - give agents a grounded draft inside every ticket
3 hr/wkest. time saved How this estimate is calculated
How it's done today

For each ticket, you identify the issue, search help articles and prior cases, check account context, write the explanation and next step, and make sure the reply uses approved language.

How AI helps

When a ticket arrives, your agent finds the relevant approved procedure, uses the ticket and allowed account context, and leaves a concise reply draft with the sources and any missing information.

Zendesk Ticket assigned customer message is complete
Your AI agent

Identifies the issue, retrieves the approved answer, applies relevant customer context, and drafts the next useful response without sending it.

Reply draft Reply draft
Source links Source links
How to set it up

Required

ZendeskIntercomHelp desk

Reads the ticket, thread, customer, product, priority, and prior handling.

Writes a private draft reply and suggested fields only.

ZendeskConfluenceKnowledge base

Reads approved procedures, policies, troubleshooting steps, and article freshness.

Recommended

SalesforceHubSpotCRM

Reads plan, account status, products, and known commitments relevant to the reply.

Optional

SlackMicrosoft TeamsChat

Writes an escalation note when the workflow cannot answer safely.

Paste the ticket thread and relevant help article into the agent. The connected version mainly removes searching and places the draft directly in the help desk.

Have current knowledge articles, response style guidance, escalation rules, prohibited promises, and 20 representative tickets with final replies and outcomes.

```text Setup prompt theme={null} Help me build a support-reply drafting workflow. For each assigned ticket, prepare a grounded reply for the agent to review. Never send it. 1. Ask for our help desk, approved knowledge sources, tone, required fields, escalation rules, and promises the agent must never make. 2. Read the full thread and identify the customer's issue, desired outcome, product, plan, steps already tried, and information still missing. 3. Retrieve the current approved procedure and cite the exact articles used. 4. Draft a concise response that acknowledges the issue, gives the next useful step in order, and asks only for information that is genuinely required. 5. Do not invent account facts, policy exceptions, refunds, timelines, or fixes. 6. Save a draft plus source links, confidence, and an escalation note if needed. 7. Test on a routine question, incomplete report, and policy-sensitive request. ```

What good looks like

The reply should answer the actual question, use a current approved source, avoid repeating steps already tried, ask only necessary questions, and make escalation obvious before an agent reads the whole thread again.

Choose your trigger

Run when a new ticket is assigned or the customer adds a message. Exclude spam, empty tickets, active incidents, legal threats, and sensitive account actions from automatic drafting.

What runs without you

Sending is never automated - the draft waits for the agent, always. Review every draft for the first 50 tickets; after that, drafting can run on every eligible ticket while agents accept, edit, or discard. Keep sampling five drafts a week for grounding and tone, because quality drift shows up in drafts nobody complained about.

Pairs well with find approved answers and draft knowledge updates - retrieval grounds the drafts, and the gaps become knowledge fixes.

Resolve routine requests

Best for a narrow set of high-volume issues with complete procedures
2.5 hr/wkest. time saved How this estimate is calculated
How it's done today

Agents repeatedly identify the same intent, verify basic eligibility, follow a known procedure, send standard instructions, and close or route the request.

How AI helps

Your agent recognizes an approved routine intent, gathers the minimum required facts, executes only allowed low-risk steps, confirms the result, and hands anything outside the procedure to a human with a clean summary.

Intercom Routine intent detected eligible for approved automation
Your AI agent

Checks eligibility, follows the exact procedure, records each action, confirms the outcome, and escalates at the first unsupported condition.

Customer resolution Customer resolution
Action log Action log
Human handoff Human handoff
How to set it up

Required

ZendeskConfluenceKnowledge base

Reads the approved intent, eligibility rules, procedure, customer message, and escalation boundaries.

ZendeskIntercomHelp desk

Reads the conversation and required customer fields.

Writes messages, status, tags, and action log within the approved flow.

Recommended

SalesforceHubSpotCRM

Reads plan and account status needed for eligibility.

Writes a resolution note when required.

Optional

SlackMicrosoft TeamsChat

Writes urgent or policy-sensitive handoffs to the owning team.

Run the same flow in agent-assist mode first: the AI proposes the steps and response while the agent clicks. Automate actions only after the intent and procedure are proven stable.

Choose one intent, document eligibility and every allowed action, define the success confirmation and escalation conditions, and collect successful, unsuccessful, ambiguous, and abusive examples.

```text Setup prompt theme={null} Help me build an automated routine-support workflow for one intent. It should resolve only the approved intent end to end and hand everything else to a human. It must never act outside the documented procedure. 1. Ask for the exact intent, approved procedure, eligibility fields, allowed actions, customer confirmations, exclusions, and escalation owner. 2. Classify the request only when the evidence meets the intent definition; otherwise hand off without taking action. 3. Gather the minimum required information and verify eligibility before acting. 4. Follow the procedure step by step, log each action and response, and stop on missing data, tool failure, policy conflict, or unexpected customer state. 5. Confirm the result in the system before telling the customer it succeeded. 6. Close only after confirmation; otherwise create a human handoff with the issue, checks, actions, result, and recommended next step. 7. Test on success, ineligible, missing-data, and tool-failure cases. ```

What good looks like

The workflow should act only on the intended request, verify eligibility and final state, keep a complete action log, and hand off quickly with useful context whenever the case leaves the documented path.

Choose your trigger

Enable only for the chosen intent and eligible products, plans, regions, and account states. Start with agent approval for every action.

What runs without you

Run it in agent-assist mode first: the AI proposes the steps, the agent clicks. Unattended execution is earned per intent - at least 95% correct classification and 100% safe escalation on replayed cases before the first automated action, and the bar resets whenever the procedure changes. Audit the action logs weekly; this is the one support workflow that touches customer accounts.

Pairs well with triage support tickets and find approved answers - triage supplies clean intents and retrieval supplies the procedure.

Find approved answers

Best for agents who know an answer exists but lose time finding it
2 hr/wkest. time saved How this estimate is calculated
How it's done today

You search several knowledge tools with slightly different words, open outdated pages, scan long articles, and compare policy versions before you can use one paragraph in the ticket.

How AI helps

Your agent searches only approved sources, returns the exact answer with a short excerpt, version and source link, and says when the knowledge base does not support an answer.

Notion Agent asks a question or opens a ticket
Your AI agent

Rewrites the issue as a search question, retrieves current authoritative passages, and returns a concise answer with freshness and source links.

Grounded answer Grounded answer
Source links Source links
Knowledge gap Knowledge gap
How to set it up

Required

ZendeskConfluenceKnowledge base

Reads approved help content, policy, product documentation, version, and ownership metadata.

Recommended

ZendeskIntercomHelp desk

Reads the ticket and product context so the search reflects the actual issue.

Writes a private answer suggestion and cited sources.

Optional

GleanNotionInternal knowledge

Reads approved runbooks and internal troubleshooting guidance.

Ask the agent a question and attach the relevant knowledge collection. Do not mix unreviewed chat history into the authoritative source set.

Have a clearly scoped approved collection, owners and review dates, archived-content rules, product and plan metadata, and 30 real questions with known answers or known gaps.

```text Setup prompt theme={null} Help me build an approved-answer retrieval workflow. For an agent question or ticket, return a concise answer supported only by our current approved knowledge. 1. Ask which collections are authoritative, how versions and products are identified, and what content is draft, archived, or restricted. 2. Convert the issue into a precise search question using product, plan, version, region, and customer goal when available. 3. Retrieve the strongest current passages and prefer the owning source over copied or older pages. 4. Return the answer, source titles and links, relevant version or date, and a short excerpt showing support. 5. If sources conflict or do not answer the question, say so and create a knowledge-gap item; never compose policy from adjacent material. 6. Keep the answer internal or as a draft for the agent. 7. Test on a direct answer, synonym-heavy question, conflict, and real gap. ```

What good looks like

The answer should use the right product and version, link to the authoritative source, avoid unsupported synthesis, and make “we do not have an approved answer” a clear and useful outcome.

Choose your trigger

Run from the help-desk sidebar, an agent question, or automatically when a ticket is assigned. Exclude draft, archived, restricted, and expired content from the search index.

What runs without you

The agent chooses what to use - retrieval only ever suggests. After 50 accurate searches, surface suggestions automatically on every eligible ticket. Review the failed searches weekly: they are your knowledge-gap backlog, and this workflow is only as good as the content it retrieves from.

Pairs well with draft support replies and draft knowledge updates - the same knowledge base powers the drafts and collects the gaps.

Review support quality

Best for reviewing more conversations against one consistent rubric
2 hr/wkest. time saved How this estimate is calculated
How it's done today

Leads sample a small number of tickets, read each thread, score a rubric, copy examples, and write coaching notes while calibration differences make scores hard to compare.

How AI helps

Your agent samples eligible conversations, scores each rubric item with quoted evidence, flags uncertain cases, and drafts coaching themes for a lead to calibrate and approve.

Zendesk Weekly QA sample eligible resolved tickets selected
Your AI agent

Applies the approved rubric, cites exact conversation evidence, and separates scoreable behavior from outcome or customer mood.

QA scorecards QA scorecards
Evidence clips Evidence clips
Coaching themes Coaching themes
How to set it up

Required

ZendeskIntercomHelp desk

Reads eligible resolved conversations, metadata, outcomes, and policy sources.

MaestroQAZendesk QASupport QA

Reads the rubric, scoring anchors, calibration examples, and sampling rules.

Writes draft scorecards and evidence.

Optional

Google DocsNotionDocs

Writes team-level themes and a calibration pack.

Export a stratified sample with complete threads and metadata. Keep agent names hidden during calibration when possible so the rubric, not reputation, drives the score.

Have a short observable rubric, scored anchor examples, sampling plan, excluded ticket types, calibration cadence, and a policy for how scores are used.

```text Setup prompt theme={null} Help me build a support-quality review workflow. Each week, prepare evidence-backed draft scorecards for a defined ticket sample. 1. Ask for the rubric, scoring anchors, eligible population, sample rules, excluded cases, calibration process, and reporting audience. 2. Select a representative sample across channels, issue types, agents, and outcomes. 3. Score only observable rubric items and cite the exact message or action for every score; mark not applicable and uncertain cases explicitly. 4. Do not infer effort, intent, or competence from response time, sentiment, or outcome alone. 5. Draft one specific coaching strength and one improvement tied to evidence. 6. Aggregate themes only after the lead approves calibrated scorecards. 7. Test against high, low, and disputed human-scored examples. ```

What good looks like

Every score should have exact evidence and a matching rubric anchor, uncertain items should reach a lead, similar behavior should score consistently, and coaching should be specific enough to practice.

Choose your trigger

Run weekly after resolved-ticket data is complete. Use stratified sampling and exclude active incidents, spam, and categories without an applicable rubric.

What runs without you

Draft scorecards can generate automatically once the agent agrees with human reviewers on at least 90% of rubric items across 50 tickets. Leads still approve every scorecard and own every coaching conversation - the agent scales the reading, not the judgment. Recalibrate monthly with disputed examples so scores stay comparable.

Pairs well with draft knowledge updates - recurring quality findings usually point at a missing or stale article.

Triage support tickets

Best for getting urgent and specialized tickets to the right queue sooner
1.5 hr/wkest. time saved How this estimate is calculated
How it's done today

Someone reads the first message, chooses category, product, language, severity, and team, checks customer status, and manually corrects tickets that landed in the wrong queue.

How AI helps

On ticket creation, your agent names and summarizes the issue, assigns approved fields, checks explicit urgency signals and account context, and routes it with a visible reason.

Zendesk Ticket created first customer message received
Your AI agent

Classifies intent, product, language, and urgency, applies customer context, and routes to the agreed queue with a concise explanation.

Ticket fields Ticket fields
Assigned queue Assigned queue
Urgent alert Urgent alert
How to set it up

Required

ZendeskIntercomHelp desk

Reads the message, channel, attachments, and routing taxonomy.

Writes title, summary, category, priority, language, tags, and queue.

Recommended

SalesforceHubSpotCRM

Reads customer tier, product, region, account owner, and open incidents.

Optional

SlackMicrosoft TeamsChat

Writes alerts for the narrowly defined urgent cases.

Run classifications as suggestions inside the intake queue. A lead can bulk-accept them while the taxonomy and examples improve.

Have a small routing taxonomy, queue owners, explicit priority rules, customer-tier fields, examples from every class, and a catch-all route for low-confidence tickets.

```text Setup prompt theme={null} Help me build a ticket-triage workflow. For every new ticket, assign approved fields and route it to the right queue. 1. Ask for the category, product, language, priority, and queue taxonomy plus definitions, examples, exclusions, and owners. 2. Read the first message and allowed account context; generate a descriptive title and one-sentence issue summary. 3. Classify only into existing labels and give the evidence for priority and route. 4. Use explicit impact, outage, security, safety, and account rules for urgency; do not rely on tone alone. 5. Route low-confidence, multi-issue, or taxonomy-gap tickets to human triage. 6. Write the fields and queue, and alert only the defined urgent owner. 7. Test on routine, urgent, multilingual, multi-issue, and ambiguous tickets. ```

What good looks like

Tickets should land in the correct actionable queue, priority should follow explicit impact rules, low-confidence cases should not be forced into a label, and the assigned agent should understand the issue from the title and summary.

Choose your trigger

Run immediately after ticket creation. Exclude spam and system notifications first, and send multi-issue or unsupported-language tickets to the catch-all queue.

What runs without you

Keep routing as suggestions for the first 100 tickets while a lead bulk-accepts. Then automate class by class - each category earns autonomy at 95% correct routing, and anything below stays suggested. Sample the automated classes weekly; taxonomy drift is invisible until a queue quietly fills with mismatches.

Pairs well with resolve routine requests and summarize conversations - clean classification is what makes safe automation and useful handoffs possible.

Summarize conversations

Best for handoffs, escalations, and long-running customer issues
1.5 hr/wkest. time saved How this estimate is calculated
How it's done today

Before taking over a case, you reread the full thread, identify the original problem, steps tried, promises, current status, and next owner, then rewrite it for an escalation or account record.

How AI helps

Your agent turns the entire thread and action log into a structured handoff with chronology, verified facts, attempts and outcomes, commitments, current blocker, and next step.

Zendesk Ticket changes owner or escalates to another team
Your AI agent

Reconstructs the issue and timeline, separates customer statements from system actions, and surfaces the current blocker and commitments.

Handoff summary Handoff summary
Account note Account note
Next actions Next actions
How to set it up

Required

ZendeskIntercomHelp desk

Reads the full thread, private notes, status changes, attachments, and action log.

Writes the structured internal summary.

Recommended

SalesforceHubSpotCRM

Reads account, product, owner, and related open cases.

Writes an escalation or account note when approved.

Optional

SlackMicrosoft TeamsChat

Writes a concise escalation message with a link to the ticket.

Export or paste the complete thread, not only the latest messages. Include the action log so attempted fixes are not inferred from conversation alone.

Have the handoff template, definitions for fact versus customer claim, required commitment fields, and examples of strong and misleading summaries.

```text Setup prompt theme={null} Help me build a support-conversation summary workflow. When a ticket changes owner or escalates, create a structured internal handoff. 1. Ask for the handoff template, required fields, destination, and which sources establish system actions and customer commitments. 2. Read the full public thread, private notes, attachments, status history, and action log in chronological order. 3. Summarize the original issue, impact, current state, relevant account context, steps attempted with outcomes, customer-provided evidence, and open blocker. 4. List commitments with owner and date only when they were explicitly recorded. 5. Keep customer claims, agent conclusions, and confirmed system facts distinct. 6. Write the handoff plus the next recommended action and source links. 7. Test on a simple handoff, repeated failed troubleshooting, and an escalation. ```

What good looks like

The new owner should understand the issue without rereading the thread, see every meaningful attempt and result, know what has been promised, and distinguish confirmed facts from customer or agent claims.

Choose your trigger

Run on owner, queue, or escalation changes and optionally after a long thread exceeds a message threshold. Exclude resolved spam and empty system tickets.

What runs without you

After 20 accurate summaries, let them write automatically on every owner change - the new owner edits instead of rereading. What stays human is trust in the content: sample handoffs weekly against the full thread, because a summary that silently drops a commitment costs a customer relationship.

Pairs well with triage support tickets - both turn a raw thread into something the next owner can act on.

Draft knowledge updates

Best for turning repeated solved issues into maintained help content
1 hr/wkest. time saved How this estimate is calculated
How it's done today

A support lead spots repeated questions, finds solved examples, confirms the current procedure with an expert, and drafts or revises an article after the gap has already created more tickets.

How AI helps

Your agent finds repeated resolved issues and failed searches, groups the evidence, compares it with current articles, and prepares a source-linked new article or change proposal for the owner.

Zendesk Knowledge gap repeats or failed search crosses threshold
Your AI agent

Collects solved examples, identifies the missing or stale instruction, drafts the smallest useful update, and routes it to the accountable owner.

Article draft Article draft
Evidence pack Evidence pack
Review task Review task
How to set it up

Required

ZendeskIntercomHelp desk

Reads resolved tickets, outcomes, searches, tags, and repeated agent workarounds.

ZendeskConfluenceKnowledge base

Reads current articles, owners, versions, analytics, and style guide.

Writes a draft article or revision, never publication.

Recommended

GleanNotionInternal knowledge

Reads approved product or policy source material.

Optional

AsanaMonday.comTasks

Writes the owner review with evidence and requested decision.

Paste a small set of resolved examples and the current article into the agent. An expert still needs to confirm that the successful support workaround is the approved product procedure.

Have a repeat threshold, solved examples, failed-search reports, article owners, product sources, style template, and a definition of what requires expert or policy approval.

```text Setup prompt theme={null} Help me build a support-to-knowledge workflow. When a knowledge gap repeats, prepare a source-linked article draft or revision. 1. Ask for the repeat threshold, eligible ticket outcomes, knowledge owners, style template, authoritative product sources, and approval rules. 2. Group resolved tickets and failed searches by the underlying customer question, preserving ticket IDs, products, versions, and successful resolution evidence. 3. Check whether an article exists, is hard to find, outdated, incomplete, or absent. 4. Draft the smallest useful change with prerequisites, ordered steps, expected result, common failure states, escalation path, and related links. 5. Do not turn an improvised workaround into policy without owner confirmation. 6. Create the draft, evidence pack, owner review task, and suggested search terms. 7. Test on a missing article, stale article, and discoverability problem. ```

What good looks like

The draft should answer a demonstrated repeated question, use approved product facts, match the right version, include a verifiable outcome and escalation path, and give the owner the exact ticket evidence behind the change.

Choose your trigger

Run weekly when repeated solved issues or failed searches cross the agreed threshold. Exclude one-off edge cases and unresolved tickets.

What runs without you

Publication stays with the knowledge owner permanently - a support workaround only becomes policy after an expert confirms it. After five accepted updates, let drafts and their review tasks generate automatically when gaps cross the threshold. Check each published change 30 days later: did the repeated tickets actually stop?

Pairs well with find approved answers and review support quality - failed searches and QA findings are its two best gap detectors.

## How to choose * Start with **draft support replies** or **find approved answers** - agents keep control, and the knowledge gaps surface immediately. * **High-volume queue?** **Triage support tickets** first, then graduate the narrowest intents into **resolve routine requests**. * **Leading the team?** **Review support quality** and **summarize conversations** scale the reading a lead cannot do alone. * **Owning the knowledge base?** **Draft knowledge updates** turns every repeated ticket into the content that prevents the next one. ## What didn't make the list (yet) Two support categories are marketed heavily and deliberately missing here: **AI voice agents** - bots that take customer calls - are absent from the defaults by design. Voice removes the safety net every workflow above relies on: there is no draft to review mid-call, and a customer trapped in a loop is your worst outcome at your busiest moment. If you go there, treat it as its own project with its own escape hatches. **Sentiment scoring as a product** - dashboards that grade how customers feel - shows up above only as a sorting signal inside triage and QA. Acting on sentiment alone punishes frustrated customers for being right; the workflows act on issue, impact, and eligibility instead. ## Frequently Asked Questions AI can find approved answers, draft replies, resolve a narrow set of routine requests, route tickets, summarize conversations, review quality, and turn repeated issues into knowledge-base drafts. Draft support replies is usually the best start. Agents keep control, the output is easy to compare with existing responses, and the workflow quickly exposes gaps in the knowledge base. Yes, for a bounded set of low-risk intents with complete approved procedures and clear escalation rules. Start in draft mode, test heavily, and expand only after resolution and escalation quality are consistently strong. At minimum it needs the current ticket and an approved knowledge source. Customer and product context from the help desk or CRM can improve the answer, but the workflow should use only the fields needed for that issue.