// Resources
Four kinds of firm sell this work in Los Angeles and they are not competing for the same job. What each one publishes about itself, with the date it was read, which California rules change the build, and when to pick somebody other than us.
Los Angeles has four kinds of firm selling this work and they are not competing for the same job. Digital product agencies that added an AI line, Goji Labs among them, build software and workflow automation and publish a typical engagement range of $50,000 to $1,000,000 on its own FAQ. Southern California software firms with offshore delivery teams, such as Pegasus One, sell engineering capacity at scale. Global consultancies with LA offices, Slalom among them, run platform and data programmes for companies that already have internal teams. Strategy and build firms, BLACKSIG SYSTEMS among them, decide what is worth automating and then build it into the software you already run and operate it afterwards. Almost nobody publishes a rate. The sorting question is not which firm is best. It is how big your company is, whether your work touches California's new automated decision rules, and who is operating the system in month seven.
Look for five things in a Los Angeles AI automation consultant, and check all five before anybody shows you a demo.
Does the firm decide, or only build? Two different purchases get sold under one name. One ends in a decision: which processes are worth automating here, in what order, and what each one costs you today. The other ends in a build against a specification somebody else wrote. Both are legitimate and they cost different amounts, so establish which one you are being quoted for before the number arrives. Our own strategy work is the first kind, and the build follows from it.
Does it write into the systems you already run? A system that reads your email and leaves the answer in a spreadsheet for somebody to re-key has not removed the work. Ask which systems it will write to by name: the CRM, the ledger, the practice management software, the field service app, the studio's production tracker. Vagueness here means integration is the part nobody has thought about, and integration is where these projects stall.
Who operates it after launch? Automations break for boring reasons. A supplier changes an invoice layout, a vendor retires an API, a form gains a field. Ask whether the firm runs the system afterwards or hands over a repository and an invoice. Systems we build live on our infrastructure and we run them from there.
Can it show you something running? Not a sandbox. A named client, a described process, a number attached. One real case beats a page of logos, and our case study with Universidad Maimonides exists so you can apply that test to us.
Does it know which California rules touch your process? This filter is new and it is sharper than it looks. A firm quoting you an automated hiring screen or a tenant scoring step without raising the California Civil Rights Council regulations or the CPPA's automated decisionmaking rules has not read them. The dates are in the California section below.
Those five sort the market better than any ranking, this one included.
The business automation consultants working with Los Angeles companies fall into four groups, listed below with a verifiable LA presence or a published LA service line, read on 2 October 2026. Every claim in the table is what the firm publishes about itself, attributed in the row, because in this category almost all public information is self reported.
| Firm | What it is | Who it is for | What it publishes about budgets |
|---|---|---|---|
| BLACKSIG SYSTEMS | AI strategy and transformation firm. We find where AI will create the most value, our engineering team builds what the plan calls for, and we run the result on our infrastructure | Owner led companies from roughly 5 to 200 staff with a repetitive process eating real hours, and mid market firms that want a decision before a build | No rate card published. Scoped on a call, against the process |
| Goji Labs | Los Angeles digital product agency founded in 2014. Says it has launched 500 or more products, supported 25 million or more users, and worked with clients who have raised over $1 billion. Runs an AI workflow automation service line alongside app, software and UX work (gojilabs.com, accessed 2 Oct 2026) | Companies that want a product built, with automation as part of the build, and funded startups shipping something new | Most engagements $50,000 to $1,000,000 on a blended rate time and materials model, with a 30 to 90 day buying cycle; fixed fees available for tightly scoped work such as an audit or a strategy sprint (its own FAQ, accessed 2 Oct 2026) |
| Pegasus One | Southern California software firm founded in 2009, with its corporate office in Fullerton and delivery centres in Tijuana, Mumbai, Delhi and Kochi. Publishes AI strategy, AI agents, generative AI, AI automation and AI governance service lines (pegasusone.com, accessed 2 Oct 2026) | Companies that want sustained engineering capacity at a blended cost and have someone internal to direct it | No rate card published |
| Slalom | Global consultancy with a Los Angeles office at 12045 Waterfront Drive in Playa Vista. Publishes an AI consulting practice and claims more than 500 AI projects in 2025 at a 3x average return (slalom.com, accessed 2 Oct 2026) | Larger companies running platform, cloud and data programmes with internal teams to work alongside | No rate card published |
| Hiring in-house | An automation engineer on payroll | Companies with years of continuous automation work queued behind the first project | Salary. Built In puts the 2026 Los Angeles average at $107,036 base plus $17,084 additional cash, $124,120 total; Glassdoor gives $126,174 average from 223 reported salaries as of July 2026, with a typical range of $98,808 to $162,640 |
Longer lists exist on DesignRush, Clutch and Built In, and most "top AI automation agencies in Los Angeles" articles are assembled from them. Those lists find you names. They are close to useless for telling a firm that decides from a development shop that added an AI page, because the entries are written by the firms themselves, and on this query several of the ranking listicles are published by agencies that appear inside their own comparison. Worth knowing before you read any of them as neutral.
Most Los Angeles AI automation consultants publish nothing about what they charge, and the one published band in the table above spans twenty times from end to end, which tells you the question is wrongly put.
Goji Labs states on its own FAQ that most engagements run $50,000 to $1,000,000 on a blended rate time and materials basis, with a 30 to 90 day buying cycle and fixed fees available for tightly scoped deliverables such as an audit or a strategy sprint, read 2 October 2026. That range is honest and it is the shape of the market. Nothing about the firm explains the difference between its two ends. The work does.
Four things move the number. How many systems the automation has to touch, because each connection is separate engineering that can break on its own. How much variation sits in the inputs, since four regular suppliers is a small build and two hundred irregular ones is not. How much has to run without a person checking each item, which is where the real cost of accuracy lives. And whether anybody operates the thing after launch, which is a monthly commitment rather than a project line.
The figure worth having before any of these conversations is your own. Count the hours the process consumes each week, multiply by loaded hourly cost, annualise it. That number is what the manual version already costs you, and it is the only thing that makes two quotes comparable. Most owners find the process they assumed was trivial is the expensive one, because its cost is spread across four people rather than sitting on one desk. The method is in how much AI automation costs.
BLACKSIG does not publish a rate. We scope against the process on a call, for the same reason a contractor does not publish a price for a kitchen.
Hiring an automation engineer in Los Angeles is cheaper than a firm only when the automation work is continuous, and for most companies the first year is not.
Built In's 2026 Los Angeles data puts the average automation engineer at $107,036 base with $17,084 in additional cash compensation, $124,120 in total, accessed 2 October 2026. Glassdoor gives $126,174 as the Los Angeles average from 223 reported salaries as of July 2026, with a typical range of $98,808 to $162,640 and senior roles averaging $181,271. Add employer taxes, benefits, recruiting and the months before anyone is productive, and one hire commits roughly a year of senior salary before the first system runs.
The harder question is what that person does in month nine, once the first three automations are live and stable. Continuous automation work exists at some companies and not at others. Where it exists, hire. Where the work is one significant build with a long maintenance tail, a hire is an expensive way to buy a project and a fragile way to buy operations, because the knowledge leaves when the person does.
There is evidence on the build versus buy question from outside the sales pitch. MIT's Project NANDA report The GenAI Divide: State of AI in Business 2025, which reviewed more than 300 publicly disclosed AI initiatives and interviewed representatives of 52 organisations, found that in its interview sample tools built with external partners reached deployment about 67% of the time against about 33% for internally built ones. That is a sample of enterprises rather than a law of nature, and it is the only published number we have found that speaks directly to the choice. The full comparison, including the case for hiring, is in automation consultant versus in-house.
Working in California changes two things, because the state changed the rules twice in the last year and both changes land on the kind of automation a growing company actually buys.
The California Civil Rights Council's regulations on automated decision systems under the Fair Employment and Housing Act took effect on 1 October 2025 and reach employers with five or more employees in the state. They make clear that an employer cannot use an automated decision system to make employment decisions that discriminate, that the employer carries responsibility even when the tool came from a third party vendor, and that records relating to automated decision systems have to be kept for four years. They also make anti-bias testing, before and after adopting a tool, an express defence against a discrimination claim, which is worth writing into a vendor agreement rather than discovering later. Mayer Brown's client alert sets out the detail. If the automation you are buying screens applicants, ranks candidates or flags employees for discipline, that regulation is a design input rather than a later review.
The California Privacy Protection Agency's automated decisionmaking technology regulations were filed with the Secretary of State on 22 September 2025 and took effect on 1 January 2026. For businesses covered by the CCPA that use such technology for significant decisions about California consumers, they require a pre-use notice describing how the technology works and what influences its output, at least two ways to opt out, and risk assessments. Businesses already using the technology have until 1 January 2027 to get notice and opt out in place. Pillsbury's summary of the new California AI laws covers these alongside the rest of the 2026 set.
California's AI rules apply only if the automation makes decisions about people, which is a sharper line than most vendors draw.
A document extraction system that reads supplier invoices, a scheduling automation that confirms appointments, a system that drafts replies for a person to send: none of those makes a significant decision about a human being, and none of them pulls the employment or automated decisionmaking obligations along with it. A screen that ranks job applicants, a tool that flags employees for discipline, a model that scores tenants or credit applicants: those do, and the compliance work rides along with the project from the first design session rather than arriving at a review in month four.
Two things to do with that. Ask any firm you are considering what it does about the Civil Rights Council regulations and the CPPA rules, by name, because a firm that has not read them will tell you quickly. And establish on the first call which side of the line your process sits on, since it changes the scope, the documentation and the four year retention question.
What California does not change is the sorting question. A dental group in Pasadena and a dental group in Ohio have the same problem with after hours calls, and the industry boards for dental and medical practices, home services, legal and accounting firms and professional firms apply the same either way. Our Los Angeles page covers what the local industry mix means for a build.
Entertainment and media lead the Los Angeles buying, and the rest of the demand looks like any large city economy.
Slalom's own Los Angeles page describes the market it serves as motion pictures, television production, video games and music, read 2 October 2026. That is a fair description of where the buying happens, and the first group has a specific shape: rights, deliverables, clearances and freelance payments produce enormous document handling and almost no two productions do it the same way.
Underneath that, the LA basin runs the largest port complex in the country, a dense logistics and apparel supply chain, an aerospace and defence manufacturing base, and tens of thousands of professional service firms and trades businesses. Those operations buy for the same reasons anybody does: intake arriving faster than people can answer it, documents that have to be read and re-keyed, follow up that nobody has time for.
Our own work runs across professional firms, home services, legal and accounting, dental and medical practices and manufacturing, and the capabilities for each are on the industry boards with a maturity label on every item marked proven, working or early. Reading those before a sales conversation is a cheap way to see what is real.
Pick someone other than BLACKSIG in four situations, and each one is a different shape of firm.
You are buying a product, not a system. Some teams want a digital product designed and built, with a roadmap, a design system and users outside the company. Goji Labs is that shape and has been since 2014. We automate how a business runs; that is a different job from shipping an app.
You want engineering capacity you direct yourself. Companies with an internal technical lead and a backlog often want engineers working to their own direction at a blended rate. Pegasus One publishes exactly that model with onshore and offshore teams. It is a reasonable purchase and it is not what we sell. We decide what is worth building, build it, and run it afterwards.
You are running an enterprise platform programme. Companies with internal data teams, a platform selection under way and a governance function need a firm structured for years of parallel workstreams. Slalom is that shape, with an LA office and a published AI practice. We are not, and a firm our size pretending otherwise is how programmes stall.
The answer is one product you can buy. If the problem is scheduling and a scheduling product solves it, buy the product. The test is whether you would change how you work to fit the tool. If you would, the tool is fine. Where that leads for trades, the best AI tools for home service businesses names the products worth buying.
What we are built for is the other case: a business running on people rather than systems, where the most valuable thing AI could do is not obvious from outside, and where somebody has to own the result rather than the deliverable.
Check three things before you sign with an AI automation firm, and none of them require a technical background.
Ask for one process end to end. Not a case study summary, a walkthrough: what arrived, what the system did with it, what a person still does, what happens when the input is wrong. Firms that operate what they build can answer in detail because they get the support calls. Firms that delivered and left tend to describe the architecture instead, which is a different answer to a different question.
Ask what it has recommended against. A firm that has never told a client to buy something off the shelf, or to fix its records before automating anything, either has not been asked hard questions or does not say no. We publish maturity labels on the industry boards so you can see that judgement before you are in a sales conversation.
Ask who watches it. Gartner's press release of 29 July 2024 predicted that at least 30% of generative AI projects would be abandoned after proof of concept by the end of 2025, naming poor data quality, inadequate risk controls, escalating costs and unclear business value. Three of those four are ownership failures rather than model failures, and they are the ones a signed operating arrangement prevents. The failure analysis in full is in why AI automation projects fail.
Directory profiles answer none of the three. They are marketing copy with a review score attached, and reading five of them produces a list of names rather than a decision.
Almost no LA firm publishes a rate. The one published band in our table, from Goji Labs, runs from $50,000 to $1,000,000 for most engagements on a blended rate time and materials model, with fixed fees available for tightly scoped work such as an audit or a strategy sprint, read on its own FAQ on 2 October 2026. What moves a quote is how many systems the work touches, how variable the inputs are, how much runs unattended, and whether anybody operates the result afterwards. Get your own number first: hours per week on the process, times loaded hourly cost, annualised. BLACKSIG scopes on a call rather than publishing a rate card.
Four stages. We map how the business runs and where the expensive work sits, decide which processes are worth automating and in what order, build the first one into the software you already use, then run it and improve it. The first stage is the paid strategy work and it produces a decision rather than a document. The build follows the decision, our engineering team does it, and the system lives on our infrastructure afterwards so the maintenance is ours rather than yours.
Judge the firm on the work rather than the address, then use locality as a tiebreaker. Almost all of this work is done remotely, including by firms with an LA office. Being in the building matters for one thing, which is the scoping conversation where the people who do the work explain what actually happens when an input is wrong. If a firm cannot be in the room for that, ask how it plans to learn the job. California's automated decision rules are a second reason to ask about local knowledge, though reading a regulation does not require an office in Playa Vista.
Specificity is not the obstacle. A process that reads the same kinds of documents, applies rules a person could write down, and produces an output somebody checks is a good candidate however unusual the business. Two things do block it: work whose rules change every time and cannot be described even by the person doing it, and work whose inputs are so inconsistent that there is nothing stable to build against. Both are worth naming early, and a firm that finds them on the call is doing its job.
It puts obligations on you, which is different from risk but worth planning for. Under the Civil Rights Council regulations effective 1 October 2025, which reach employers with five or more employees in California, an employer carries responsibility for a discriminatory automated decision even when a vendor supplied the tool, and has to keep automated decision system records for four years. Those regulations also make anti-bias testing before and after adoption an express defence, which is a reasonable thing to require of a vendor in writing. Under the CPPA regulations effective 1 January 2026, covered businesses using automated decisionmaking technology for significant decisions owe a pre-use notice and at least two opt out routes, with until 1 January 2027 to have notice and opt out in place for existing uses.
Quarterly. Every figure carries the date it was read, because service lines, published budget bands and California compliance dates in this category change within months. One figure on this page moved between the draft and publication: Glassdoor's Los Angeles average, which is recalculated continuously. If something here no longer matches what a firm publishes, the page is wrong and we want to know.
We map how your business actually runs, decide where AI is worth using, then our engineering team builds what the plan calls for and we run it from there. We own the outcome, not the deliverable.