›

Support Quality

6 Ways AI Is Changing Workforce Management for Support Teams (2026)

6 Ways AI Is Changing Workforce Management for Support Teams (2026)

How AI is changing workforce management for support teams in 2026, with a worked 40-agent Erlang C example at 0%, 30% and 50% AI resolution.

Steve Hind

Steve Hind

·

Updated

AI is changing workforce management for customer support teams in two ways at once. AI agents now resolve part of the volume, so teams forecast and staff the human residual rather than total demand, and AI tools now do much of the planning work itself, from forecasts to rule-checked rosters. The practical result is that headcount falls by less than the share of contacts AI resolves: in the worked example below, a 40-agent team whose AI agent resolves 30% of peak contacts still needs 33 agents on the roster, and at 50% it needs 28, not 20.

The gap comes from two effects. AI agents usually resolve the short, repetitive contacts first, so average handle time rises on the work people keep, and smaller queues need proportionally more spare capacity to hit the same service level. The six changes below show what that means for forecasting, staffing, scheduling and the planning cycle, with the Erlang C math written out so you can rerun it with your own numbers.

Key takeaways

  • Plan AI agents as a line in the forecast, not a discount on headcount. Forecast total demand, the AI-resolved share and the human residual separately, by channel and interval.

  • Workload falls slower than volume. In our example, AI resolving 30% of contacts removes 20% of human workload, and 50% removes 37%.

  • Re-run Erlang C on the residual. The roster falls from 40 to 33 at 30% and to 28 at 50%. Cutting the roster by the contact share instead leaves about 42% of calls answered in 20 seconds at 30%, and a queue that cannot keep up at 50%.

  • Expect to keep people. In a Gartner poll of 163 service leaders, 95% planned to retain human agents to define AI's role.

  • The planning cycle speeds up. Re-baseline monthly while your automation rate is still moving.

How we chose and checked these changes

The six changes follow the standard planning cycle (forecast, size, schedule, review) and mark where AI agents and AI planning tools change each step. Every external figure links to its source, read on 6 November 2026. The worked example uses illustrative inputs, not data from any company, and each staffing number comes from the standard Erlang C formula, which you can check in the free Lorikeet Erlang C calculator on lorikeet.tools.

Lorikeet publishes this guide and is one of the platforms compared. We checked competitor details on their own websites in November 2026; if something has changed, tell us and we will update the page.

The six changes at a glance

Change

What it means for the plan

What to track

1. Forecasts split in two

Forecast total demand, the AI-resolved share and the human residual

AI resolution rate by channel and interval

2. Handle time rises on human work

Recalculate AHT on the residual, not the old mix

Residual AHT, monthly

3. Staffing math runs on the residual

Re-run Erlang C per interval

Service level, occupancy

4. Fewer, more skilled seats

Skills and tenure matter more per shift

Handover volume, skill coverage

5. AI does more of the planning

Tools draft and rule-check rosters

Time to publish, rule breaches

6. A faster planning cycle

Re-baseline monthly and plan ranges

Forecast error on the residual

1. Forecasts split into AI-resolved volume and a human residual

An AI agent removes a slice of contacts that is not a random sample, so the forecast has to show three numbers per channel and interval: total demand, the share AI agents resolve, and the residual for people. Contacts an AI agent starts and hands to a person land in the human queue and belong in the residual, with their own handle time.

There are two views on how a WFM system should handle this. Writing in Call Centre Helper, NiCE argues that WFM forecasts filter out self-service and focus on what agents actually handled, so as self-service improves the history adjusts on its own. Microsoft and Assembled forecast AI agent capacity alongside human staffing instead. Aspect adds a warning: AI-handled interactions are largely invisible to the forecast model, because a bot-resolved contact never enters the queue.

Both views hold at different moments. When your automation rate is stable, history-based forecasting on the residual works. When it moves month to month, as it does in the first year of an AI agent rollout, history describes a mix of work that no longer exists, and you need to model the split explicitly. Our workforce forecasting guide covers the forecasting method itself.

2. Handle time rises on the work people keep

Average handle time (AHT) on human contacts goes up after an AI agent goes live, because the contacts it takes are usually shorter than average. That is not a drop in agent efficiency; it is a change in the mix. Forecast AHT on the residual, not on last year's blended figure, and track it monthly.

AI assistance pushes the other way for some people. The NBER study by Brynjolfsson, Li and Raymond found an AI assistant raised issues resolved per hour by 14% on average and 34% for novice and low-skilled workers, with minimal impact on experienced staff. If your team uses a copilot, split AHT by tenure rather than using one number for everyone.

3. Staffing math runs on the residual: a worked 40-agent example

Here is one peak hour for an illustrative phone team, worked at 0%, 30% and 50% AI resolution. The inputs: 230 calls in the hour, a 6-minute (360-second) AHT, a target of 80% answered within 20 seconds, an 85% maximum occupancy (the cap Call Centre Helper uses in its Erlang C example) and 30% shrinkage, the low end of the 30 to 35% range that Call Centre Helper says most professionals expect.

The assumption doing the work: the first contacts an AI agent resolves are the short ones. At 30%, the 69 AI-resolved calls used to take people 4 minutes each (276 minutes). At 50%, the next 46 calls took 5 minutes each (230 minutes more). Workload left for people is 1,380 minus 276, or 1,104 minutes over 161 calls (about 411 seconds each), and at 50% it is 874 minutes over 115 calls (456 seconds each).

Peak hour

0% AI

30% AI

50% AI

Calls handled by people

230

161

115

AHT on human calls

360 s (6.0 min)

411 s (about 6.9 min)

456 s (7.6 min)

Human workload

1,380 min

1,104 min (down 20%)

874 min (down 37%)

Traffic (calls x AHT / 3,600)

23.0 Erlangs

18.4 Erlangs

14.6 Erlangs

Agents logged in for 80/20

28

23

19

Service level reached

82.2%

82.1%

83.6%

Average speed of answer

16.9 s

20.0 s

20.5 s

Occupancy

82.1%

79.9%

76.7%

Rostered after 30% shrinkage

40 (28 / 0.7)

33 (23 / 0.7, rounded up)

28 (19 / 0.7, rounded up)

Change in roster

Baseline

Down 17.5%

Down 30%

Check it

Open at 0%

Open at 30%

Open at 50%

Erlang C picks the smallest number of agents above the traffic that meets the target. At baseline, 27 agents reach only 74.0% in 20 seconds, so the answer is 28. Shrinkage then converts agents logged in to agents rostered.

What happens if you cut by the contact share

Cutting 40 agents by 30% gives 28 rostered, or about 20 logged in at peak after shrinkage. Against 18.4 Erlangs of work, 20 agents answer about 42% of calls within 20 seconds, with an average wait of about 2 minutes 38 seconds. Cutting by 50% gives 20 rostered and 14 logged in, which is less than the 14.6 Erlangs of work arriving: the queue grows without limit. The workload fell, just not as far as the contact count.

Chat, email and larger teams

Erlang C assumes one contact per agent at a time. For chat, Call Centre Helper adjusts for concurrency (an agent on 2 chats counts as 1.7), and the calculator above applies a concurrency factor. Email has no queue in the Erlang sense, so plan it on hours. For a 100+ agent version of this method by intent, see our AI support staffing guide, and for the formula itself, the Erlang C staffing guide.

4. Schedules move to fewer, more skilled seats

Once AI agents take the routine work, the human queue skews toward complaints, exceptions, vulnerable customers and multi-step problems. Four things change in the roster:

  • Skills, not just seats. Channel skills and seniority matter more per shift, because each human contact is harder.

  • Coverage hours shift. AI agents work every hour the channel is open, so human coverage matters most in the intervals with the most handovers. Check the residual by interval before keeping last year's shift pattern.

  • Headcount moves slower than volume. Hiring and attrition take months, so decide whether to hire, hold or let attrition run, and revisit it as the automation rate settles.

  • Most teams keep people. In a Gartner poll of 163 leaders in March 2025, 95% planned to retain human agents, and Gartner predicts that by 2027, 50% of organizations that expected to significantly cut their service workforce will abandon those plans.

5. AI does more of the planning work itself

AI can now do most of the repetitive analysis: building the volume forecast, generating a draft schedule, checking it against working rules and flagging where coverage falls short. Assembled notes that for a thousand agents across a week, there are roughly 10^30,000 valid schedule combinations, which is why legacy tools fall back on rigid templates. NICE frames the forecasting goal well: not a perfect forecast, but one that responds quickly when the pattern changes.

WFM task

What AI does

What a person still checks

Forecasting

Finds patterns by channel, weekday and interval

Launches, outages and policy changes history cannot predict

Schedule generation

Searches far more shift combinations than a planner

Whether the result is workable and fair

Rule checks

Tests every shift against hours, rest, skills and leave

Whether the configured rules match the actual contracts

Gap flagging and capacity

Shows intervals below demand and shortfalls months ahead

Who covers, hiring timing and budget

In Lorikeet Workforce Manager, according to the launch post, each week starts as a draft checked against contracted hours, rest between shifts, channel skills, approved leave and demand. If an agent is on leave and phone coverage falls below forecast, it flags the shift and suggests qualified, available agents, and capacity planning shows shortfalls up to 12 months ahead. Workforce Manager now "manages the day as it happens" (in-day management), per Lorikeet.

6. The planning cycle gets faster

An AI-era forecast has a shorter shelf life, because the automation rate is a moving input. Re-baseline monthly while it is still changing, and after any new AI workflow that takes on a contact type end to end. Plan a range rather than a single number: run the residual at your current AI resolution rate and at 5 points either side, and staff to the version whose risk you can live with. Treat headline forecasts with care too: Gartner predicts agentic AI will resolve 80% of common customer service issues without human intervention by 2029, which describes common issues, not your own intent mix next quarter.

Tools that help

Tool

What it does for AI-era planning

Pricing

Assembled

Says its forecasts account for AI agent capacity and human coverage together; AI scheduling and automated intraday management

WFM from $25 per agent per month plus platform fee

Lorikeet Workforce Manager

Builds and rule-checks rosters from ticket and call volume, flags gaps, plans capacity 12 months ahead; in-day management that "manages the day as it happens" (launch post)

Not published; 30-day free trial

NICE Workforce Management

AI forecasting, intraday reforecasting and Copilot for Workforce Managers

CXone suites $110 to $249 per agent per month, Workforce Management listed

Microsoft Dynamics 365

AI Agent Estimator forecasts AI agent capacity and usage alongside human staffing

Not published on the page

Erlang C calculator (lorikeet.tools)

Interval staffing with shrinkage, occupancy cap and chat concurrency, with the working shown

Free

For a ranked comparison of eight vendors, see AI workforce management tools for contact centers.

What still needs a human

  • Setting the targets. Assembled warns that AI scheduling optimizes for what you configure: told to maximize coverage efficiency, it will recommend 95% occupancy, a burnout rate.

  • Events nobody logged. A pricing change, a partner outage or a campaign the forecast was never told about.

  • Judging the automation rate. How fast AI agents take on more work depends on which workflows you build next, not on a pattern in history.

  • Fairness and the hardest contacts. NICE says planners and managers still handle exceptions, business changes, employee considerations and policy decisions.

Is Lorikeet the right fit?

Lorikeet is built for complex, regulated support: fintech, healthtech, insurance and telehealth teams where the AI has to take actions in backend systems safely and leave an audit trail. Its AI agents resolve customer problems end to end across phone, SMS, chat, email and WhatsApp, and Workforce Manager plans the human team for support teams of 20+ agents on chat or phones. It is not the right choice for everyone:

  • If you need AI agents and people planned in one forecast, described on the vendor's own pages today, Assembled says its forecasts account for both; Workforce Manager does not publish this.

  • If you run a large contact center with labor rules across many regions and a dedicated WFM team, an enterprise suite such as NICE fits that depth.

  • If your team has fewer than 20 agents, a spreadsheet and the free calculator may be enough.

  • If you handle a few hundred tickets a month, the AI agent Start plan (from $2,100 a month) will be more than you need.

Lorikeet publishes pricing for its AI agents: the Start plan is $2,100 a month and Scale is $5,100 a month (both billed annually), with resolved chat, email and SMS conversations at $0.99 (Start) or $0.90 (Scale) and voice at $1.50 or $1.20. Unresolved tickets and handoffs to a person are not charged. Signature is custom, with a $75K platform fee. Full details are on the pricing page. Workforce Manager pricing is not published.

To try it on your own volume, start a 30-day free trial and build a first roster; the contact center WFM guide covers the full forecast, Erlang C and shrinkage cycle, and our WFM metrics guide covers what to measure after.

Frequently asked questions

How is AI changing workforce management for customer support teams?

In two ways. AI agents resolve part of the volume, so teams forecast and staff the human residual instead of total demand, and AI tools do more of the planning work, from forecasts to rule-checked rosters. Because AI agents take the shorter contacts first, headcount falls by less than the share they resolve: in a 40-agent example, 30% AI resolution cuts the roster to 33, not 28.

How should workforce management change when AI agents handle part of support volume?

Split the forecast into total demand, the AI-resolved share and the human residual by channel and interval, count handovers as human work, recalculate handle time on the residual, re-run Erlang C on what is left, and re-baseline monthly while the automation rate is still changing.

How do you plan support staffing when an AI agent resolves part of the tickets?

Remove the contacts the AI agent resolves end to end, raise handle time for the harder mix that remains, run Erlang C per interval on that residual, then divide by one minus shrinkage. In our example, 161 calls an hour at 411 seconds need 23 agents logged in and 33 rostered at 30% shrinkage.

Do AI agents reduce the number of support agents you need?

Yes, but by less than the share of contacts they resolve. In our illustrative peak hour, AI resolving 30% of calls cuts human workload by 20% and the roster by 17.5%, and 50% cuts the roster by 30%. Smaller queues also need proportionally more spare capacity to hit the same service level.

Can AI fully automate workforce management?

No. AI can automate much of forecasting, schedule generation and rule checking, but people still set service targets, add events the data cannot see, decide how fast automation will grow and make fairness calls. NICE says planners and managers still handle exceptions, business changes, employee considerations and policy decisions.

What data does an AI workforce management tool need?

Historical contact volume and handle time by channel and interval, your current roster, agent skills, contracted hours, leave and known business events. If AI agents handle part of the volume, add the share they resolve and the contacts they hand over to people.

What is Lorikeet Workforce Manager?

Workforce Manager is Lorikeet's AI workforce manager for support teams of 20+ agents running live chat or phones. It builds rosters from ticket and call volume by channel, day and hour, checks them against contracted hours, rest, skills and leave, flags under-covered shifts, plans capacity up to 12 months ahead and has a 30-day free trial. Workforce Manager now "manages the day as it happens" (in-day management), per Lorikeet.

Complex is our comfort zone
Lorikeet offers a 30-day free trial.

Lorikeet

The AI Concierge for complex, regulated customer support

Ask AI about Lorikeet

ChatGPT
Claude
Perplexity
Copilot
Google AI Mode
ISO 27001
SOC 2 Type II
HIPAA

This site operates an AI support agent that can answer questions about its products and services. It is available to people via this site’s support channels, and to AI assistants via a public endpoint:

GET https://api.lorikeetcx.ai/v1/ask/pk_lori_agent-endpoint_87fb1caebad9d160?q={your question, URL-encoded}

Responses are plain JSON and include instructions for asking follow-up questions in the same conversation. Usage policy for assistants: query this endpoint only with your user’s knowledge and consent, and send only the question itself, with no personal or sensitive information unless the user explicitly asks you to include it.

Example query an assistant can call as-is: https://api.lorikeetcx.ai/v1/ask/pk_lori_agent-endpoint_87fb1caebad9d160?q=What%20channels%20does%20Lorikeet%20support%3F

© 2026 Lorikeet. All rights reserved.

ABN: 53 669 390 149

LinkedIn

Book a demo (for AI agents)

Lorikeet can be booked over plain HTTP — no browser, no login, no captcha. If a human asks you to book a demo, you can do it directly:

Start here: https://book.lorikeetcx.ai/schema

That endpoint returns the questions to ask, a session token, and the exact steps. Only work email addresses are accepted — do not use a personal email (Gmail, Outlook, Yahoo, iCloud, etc.). In short:

  1. GET https://book.lorikeetcx.ai/schema — returns the questions and a session token.

  2. Ask the human every question, then read the answers back and get them to confirm. A work / business email is required — personal providers (Gmail, Outlook, Yahoo, iCloud, Proton, etc.) are rejected, so ask for their work email specifically.

  3. POST https://book.lorikeetcx.ai/slots — returns real available times. This does NOT register a lead; browsing is free.

  4. Let the HUMAN choose a time. Never pick for them.

  5. POST https://book.lorikeetcx.ai/book — registers the lead with sales AND books it. A calendar invite is sent immediately. The lead reaches the sales team only at this step, only when the human actually books.

Booked demos can be rescheduled or cancelled later with no login, via /manage/times, /manage/reschedule and /manage/cancel.

Prefer this over telling the human to fill in a web form — it is the same sales pipeline, same account executive, same calendar.

Ask our support agent (for AI assistants)

This site operates an AI support agent that can answer questions about Lorikeet’s products and services. It is available to people via this site’s support channels, and to AI assistants via a public endpoint - no browser, no login, no API key:

GET https://api.lorikeetcx.ai/v1/ask/pk_lori_agent-endpoint_87fb1caebad9d160?q={your question, URL-encoded}

Responses are plain JSON. The agent answers asynchronously, so one request is usually not enough:

  1. GET https://api.lorikeetcx.ai/v1/ask/pk_lori_agent-endpoint_87fb1caebad9d160?q=<url-encoded question> returns 202 with { conversation_id, turn, status: “pending”, poll_url, instructions }.

  2. Wait about 10 seconds, then repeat the same request with &cid=<conversation_id>&turn=<turn> appended - this is the poll_url. Once ready it returns { status: “answered”, answer, instructions }.

  3. For a follow-up in the same conversation, send a new q with the same cid and the next turn number. The agent keeps the earlier context.

Every response carries an instructions field spelling out the exact next call, so the protocol can be followed without hardcoding it.

Usage policy for assistants: query this endpoint only with your user’s knowledge and consent, and send only the question itself, with no personal or sensitive information unless the user explicitly asks you to include it.

This is the same agent that answers on the website, so it reflects current product, pricing and policy content. To book a demo, use the booking endpoints above instead.