Bland AI vs LLDTEK: pricing, plan limits and booking for home services

Bland AI is an enterprise voice platform for teams that build and run their own phone agents. Its Start and Build plans charge $0.14 or $0.12 per minute of talk time. The features a home-service office relies on most, such as text messages, warm transfers and the appointment scheduling node, are listed on the custom-priced Enterprise plan. LLDTEK sets up booking agents beside the job board a company already runs and quotes after a demo. An HVAC or plumbing company with a developer and a simple calendar can make Bland work. One without a developer will usually find a managed product closer to the job.

LLDTEK publishes this site and sells one of the two products compared here, so weigh the LLDTEK sections with that in mind.

What Bland AI is built for

Bland describes itself as "an enterprise voice AI platform" for "the calls that can't go wrong". The customer stories on its homepage come from regulated work, including the Idaho Housing and Finance Association and a legal hotline, and the page says more than 230 regulated companies run on Bland. Nobody outside Bland can check that count, but it shows who the product is sold to.

Bland does sell to home-service companies. Its industry page for home services promises "24/7 answering, booking, dispatch, and follow-up for HVAC, plumbing, electrical, and cleaning", says it schedules "by technician availability and service area", and quotes $0.11 to $0.14 a minute. The page offers a sales call or a self-serve start, but on the pricing page the Appointment Scheduling Node, texting and warm transfers sit on Enterprise.

Two engineering choices set Bland apart from most voice-agent tools. The homepage says it is "the only voice AI that runs entirely on its own models", on Bland's own infrastructure, and that self-hosted and on-premises deployments are available "for the most sensitive workloads". Enterprise customers get isolated infrastructure and choose when to adopt each release, with canary deployments for trying a new version on live traffic first. On compliance, Bland says it holds independently audited SOC 2 Type I and Type II reports and self-attests to PCI DSS, and its pricing page lists a business associate agreement, single sign-on and data residency under Enterprise.

Those are real strengths, bought for a different buyer. Isolated infrastructure and canary releases protect a call centre that handles mortgage questions. A two-branch heating and plumbing company mostly needs its agent to know which technician holds the gas certification and which ZIP codes the north van covers.

Building an agent has become easier than Bland's developer reputation suggests. In the agent builder you write instructions in plain language, add tools and knowledge, set guardrails, run simulated test calls, and promote a version from staging to production. Bland's pricing page says most teams deploy a first agent "within a day" on the self-serve plans, while Enterprise deployments follow a 28-day framework with Bland's engineers. The homepage puts most production agents at two to six weeks. Read together, an agent that answers the phone takes a day, and an agent you trust with a July dispatch board takes weeks.

Bland AI pricing, plan by plan

These figures come from Bland's pricing page and its billing documentation as of November 2026. Bland's billing docs date the current per-minute rates to December 5, 2025, so confirm them before you sign.

Start comes with "2 credits + an inbound number ($15/mo value)", and Bland's docs put that free credit at $2, about 14 minutes at Start's rate. The talk-time rate covers the language model, speech-to-text and text-to-speech, and the page's tagline begins "No token charges." Telephony is separate: it is "billed separately, on your own carrier or Bland's at pass-through cost", and teams that bring their own telephony pay no transfer fee. The page sets no contract term for Build and no minimum monthly spend.

The plan comparison on that page is where a service business should look hardest. SMS and web chat, in-call texts, the Appointment Scheduling Node, and warm and live transfers are all marked as included on Enterprise and not on Start or Build.

Bland's billing documentation fills in the smaller charges:

• Call time is prorated to the second, and voicemail is billed as ordinary call time.
• Each outbound attempt and each failed call on Bland's telephony carries a minimum charge of $0.015.
• Where SMS is available, each message costs $0.02, inbound or outbound.
• Calls over a plan's daily or hourly cap are rejected rather than charged.
• The docs also list a Scale plan at $499 a month with $0.11 talk time, which the pricing page no longer shows, and say the flat rate before December 5, 2025 was $0.09 a minute. That is why older reviews quote a lower price.

The same docs describe an Agent Phone plan at $19.99 a month after a $9.99 first month, or $150 a year. Bland built it for bots and autonomous agents that need their own number. It includes voice minutes and texts to the US and Canada, unlimited within fair-use caps, and one local number. It allows one concurrent call across the whole account, 50 calls a day and 20 an hour, and Bland says it is "built for one agent doing one thing at a time". For an owner-operator with one van, that is cheap. The catch is the single line: while the agent is on one call, the docs say new inbound callers hear a busy message.

What a two-branch HVAC company would pay

Take a heating, cooling and plumbing company with two branches that sends overflow and after-hours calls to an agent. In an ordinary month the agent takes about 1,200 calls across both branches. Each averages four minutes from pickup to hang-up, including any time after a transfer, because callers describe the fault, give an address and agree an arrival window. One call in five is urgent enough to transfer to the on-call technician or the office, with two minutes of transfer time each.

On rates alone, Start wins by $198.20 a month. Build only becomes cheaper when the $0.02 saved on each talk minute covers the $299 fee, which takes 14,950 minutes, roughly 3,700 four-minute calls a month, ignoring the smaller transfer saving. At about 125 calls a day, Start's cap would already have ruled it out.

The cap decides it instead. An ordinary month averages about 40 agent calls a day, well under Start's 100. Then the first hot week of summer arrives, air conditioners fail across town, and the agent takes 160 calls on a single day. If inbound calls count toward Start's daily cap, as Bland's billing docs imply, about 60 of those callers would be rejected by the platform on the day the company most wants the phone answered. Paying $198.20 a month more for Build buys headroom the company may use only a few weeks a year, so ask Bland whether you can move between plans by season.

Three costs sit outside both totals:

1. Telephony. Start includes one inbound number. If each branch keeps its own number, the company pays its carrier or Bland's pass-through rate for the rest, and the pricing page does not list that rate.
2. The booking connection. Someone has to build the tool that reads open windows from the job board and writes the job back. The next section explains what that involves.
3. Texts and warm transfers. If the company needs either, Start and Build do not include them, and the Enterprise price is a quote.

To compare any of these bills fairly, divide the monthly total by jobs that actually landed on the board, not by calls answered. A caller who said "I'll get another quote" used the same talk time as one who booked a furnace replacement estimate. The AI receptionist cost guide sets Bland's billing model beside per-call and flat-fee products.

Where the booking actually happens

Bland's integration tools docs walk through Cal.com, Salesforce and Slack, and the Cal.com connection offers a booking action. Its integrations list adds Calendly, HubSpot and automation tools such as Zapier and Make. Neither names ServiceTitan, Jobber or Housecall Pro. For every other system, Bland points builders to a webhook tool that calls an HTTP endpoint during the conversation. It runs GET and POST requests only, and it is set up in the builder by filling in the URL, headers and request body, or by importing a working cURL command.

The Scheduling Node shows the pattern Bland expects. Your endpoint returns "a flat array of dates in ISO format", the agent offers those times, the caller picks one, and a later webhook writes the booking. If a caller wants to stop partway, the builder adds that exit condition, and the doc covers picking a new time only, not moving or cancelling an existing job. The pricing page lists this node under Enterprise.

That flat list of times is where a service company's real rules can drop out. A dispatch board knows more than which hours are empty. It knows that a boiler repair needs a gas-certified technician, that a two-hour arrival window has to include the drive from the last job, that the south branch does not cover the far side of the river, and that a maintenance-plan member may have been promised a faster visit. None of that travels in a list of dates unless the developer's endpoint applies it before it answers.

So the substantial part of a Bland build for a home-service company is that endpoint. It has to ask ServiceTitan, Jobber, Housecall Pro or whichever board the company runs for windows that fit the job type, the technician and the address. It then has to write the job back and cope when the write fails. Whether that is possible depends on what the job board's API allows. Housecall Pro, for example, offers Zapier and its open API only on its MAX plan. Whether the connection keeps working depends on who maintains it. The Vapi comparison walks through the same build-it-yourself question on another developer platform.

Texts, transfers and a technician under a sink

Home-service customers text when they cannot talk: "can the tech come after 3?" or "gate code is 4471". Bland's SMS documentation says SMS is available on Enterprise plans and the Agent Phone plan, and that every US number used for SMS must be registered to an A2P (application-to-person) messaging campaign. On Start and Build, that text has nowhere to go inside Bland.

Transfers matter more here than in most appointment businesses, because some calls are genuinely urgent. Start and Build price transfer minutes, so a call can be passed to a phone number. The warm transfer, where Bland's agent places a second call to brief a staff member before connecting the caller, is described in Bland's docs as an Enterprise feature, though one note on the same page also lists the Agent Phone plan. If the person does not pick up, the builder chooses between ending the call with a message that the agent is unavailable or sending the caller back into the conversation flow.

The on-call technician is often in a crawlspace or driving between jobs. Whoever builds on Bland has to decide what a caller with no heat in January hears when that phone keeps ringing, and who sees the request afterwards. LLDTEK's home services page describes handing the thread, with the address, to the person the shop names, so nobody makes the customer repeat the fault. Bland's warm transfer also briefs the person it reaches. With either, ask what happens when nobody answers.

Outbound at scale is a different job from rebooking

Bland is built to dial lists, and this is where it is strongest for a service company. Its batch calls take a CSV of recipients and staff every call with one agent. Numbers on your do-not-call list are dropped at dial time, every number needs a country code, and a file with one bad row is rejected whole. Each outbound attempt on Bland's telephony carries a $0.015 minimum charge, and voicemail is billed as talk time. A company with two thousand maintenance-plan members to call before cooling season gets real use from that.

The smaller jobs depend on timing rather than volume. A cancelled 9am window needs a quick text to a nearby customer on the waitlist who can actually take it, before the van sits idle. The customer whose furnace was serviced today should have next year's visit offered before the technician leaves. Both are usually texts, which takes the company back to Bland's Enterprise plan. Automated calls and texts to customers also carry consent rules, and Bland's batch docs do not address consent, so check those rules for your market before any campaign.

Bland's platform keeps moving, and someone has to follow it

In September 2026 Bland updated its agent builder, and its migration guide calls Agents "the successor to conversational pathways". Accounts created after September 18, 2026 use Agents only. Bland asked for production agents to be migrated by November 15, 2026, after which pathways still handling traffic "are migrated for you".

Bland has handled this carefully. Migration creates new agents without touching the old pathways, numbers or traffic, and calls move only when you promote the new version. Enterprise accounts have a Bland partner run the migration. On a self-serve account, though, somebody at the company, or the agency it hired, has to test the migrated agent against the real dispatch board before promoting it. A platform change like this one is the maintenance a per-minute price does not show.

What LLDTEK does differently

LLDTEK is an AI agent platform for appointment businesses, including home and local services. Its five agents cover the front desk, cancelled-job recovery, rebooking, reviews and live availability, and they share one conversation thread across voice, SMS and the web. LLDTEK's home services page says the agent offers a window from the times the company sets, beside the job board it already runs, and that the demo confirms which systems it reads open windows from and books into. When a job drops, Recovery texts the waitlist for that window. The Reviews agent sends one Google review request after the job is done.

LLDTEK does not ask the company to build anything. The How it works page sets out five steps: connect the calendar or job board, configure rules and handoff paths, test real scenarios against the booking data, launch one agreed job beside the team, then review live conversations. The integrations page says LLDTEK connects to more than 100 booking, POS, calendar and messaging systems, with guides for ServiceTitan, Jobber and Housecall Pro among others. For ServiceTitan, Jobber and Housecall Pro, LLDTEK's guides say the exact actions inside the board are agreed on the demo.

The trade-offs are real. LLDTEK publishes no price list and quotes only after a thirty-minute demo, and it turns down teams that want to build and prompt agents themselves. Billing starts only after the team has watched a live day with LLDTEK. Bland lets anyone open an account and price the minutes tonight.

Bland AI and LLDTEK side by side

When Bland AI is the better choice

Bland fits a company that has a developer or agency committed to the project, and a job board with an API that person knows well, or bookings already in Cal.com.

It fits when the main job is calling out at volume, such as reaching every maintenance-plan member before the season turns, where batch calling is built for exactly that.

It fits a larger operator that needs what Enterprise sells: isolated infrastructure, canary releases on live traffic, compliance paperwork and engineers from Bland on a shared Slack channel.

It also fits anyone who wants to price usage tonight. Bland publishes its rates and lets you start free.

When LLDTEK fits better

LLDTEK fits a company whose jobs live on a board it is not leaving, and which wants the agent working beside that board, with the exact read and booking actions confirmed on the demo.

It fits when customers text as much as they call, and the reply has to land in the same conversation the call started, without moving to an enterprise contract.

It fits when the lost work goes beyond the phone: the cancelled window nobody refilled, the customer whose next service was never booked, the review nobody asked for.

And it fits when nobody in the office wants to maintain an agent through a platform change. If every call already gets answered, neither product is needed, and keeping the phone as it is costs nothing.

Check your job board first. ServiceTitan's Virtual Agent, Jobber's Receptionist and Housecall Pro's CSR AI each book onto their own board, and LLDTEK's ServiceTitan page says a shop already running Virtual Agent may not need another agent for the phones.

Questions to ask before you sign either one

• Which plan includes the features we need? Name texts, warm transfer and booking explicitly, and get the plan in writing.
• What happens to the 101st call of the day? Ask what the caller hears when any cap or concurrency limit is reached, and whether you can change plans for the busy season.
• Where does the booking write, and who built that connection? Make a test call that books, moves and cancels a real job on your own board, then check that the window reopened.
• What does an urgent caller hear when the on-call phone goes unanswered? Try it while the technician is on a job.
• Who updates the agent when the price book, the service area or the platform changes? Get a name, not a team.
• What does telephony cost for each branch's number? Add it to the minutes before comparing.

The field service guide to missed calls covers the dispatch-board checks in more depth, whichever vendor you test.

Which one fits a home-service company

Bland AI is a capable platform built first for regulated call centres, outbound programmes and teams with engineers, and it also markets voice agents to HVAC, plumbing, electrical and cleaning companies. A home-service company can use it, and its batch calling suits seasonal outreach well. The parts of an office that matter most day to day, such as texts, briefed transfers and booking against the real dispatch board, either sit on the Enterprise plan or depend on a connection someone has to build and maintain. Count your busiest day's calls, name the job board you will not leave, and decide who would own the agent. With a builder, a simple calendar and mostly outbound work, start with Bland's free plan. Without them, look at managed options first.