Mo. Abubakar Lala (@theablelala) header
Mo. Abubakar Lala (@theablelala) profile picture

COO, RS Gonzales (US marketing agency) | Founder & CEO, Soltiks AI agents with human review for agencies In Accounting Today, Techstrong | Writes Mo. Talks

Washington, DC · Karachi
Joined September 2017
Photo by Mo. Abubakar Lala (@theablelala)Photo by Mo. Abubakar Lala (@theablelala)Photo by Mo. Abubakar Lala (@theablelala)Photo by Mo. Abubakar Lala (@theablelala)Photo by Mo. Abubakar Lala (@theablelala)Photo by Mo. Abubakar Lala (@theablelala)Photo by Mo. Abubakar Lala (@theablelala)Photo by Mo. Abubakar Lala (@theablelala)Photo by Mo. Abubakar Lala (@theablelala)Photo by Mo. Abubakar Lala (@theablelala)
I'm Mo. I run a US marketing agency's operations on AI agents, and a named person signs off all the work. I've made 11 free kits from that system: agent setup, prompting, memory, multi-model QA, guardrails, motion graphics. Follow me and comment KITS, and I'll DM you the list.
89
135
2
1,997
372,704
Speaking at @Cloudways in Karachi tomorrow: "Agents Draft, People Decide: Stop Selling Hours, Start Selling Sign-Off." How to run client work on AI agents with a person signing off, and sell that. Fri 20 Nov, 5:30 PM PKT luma.com/gy5aa65c
13
Two heating and air markets, same trade, same day. In suburban Pennsylvania, contractor homepages held 36 of 70 top-10 places. In Central Florida, contractor pages held 3 of 56. Opposite rules. So we build rules from each market's winners: Rulesourcing.
1
3
31
UiPath's CEO is right: chain enough probabilistic steps and errors compound. 90% per step over 5 steps is 59%. So don't pick scripts or agents. Split the jobs: the model judges, scripts execute, and a named person signs off before anything reaches a client.
1
4
63
Your AI invoice shows what you paid, not what you got. In Deloitte's Q3 survey, 53% of CFOs review AI costs after the fact. Put seat yield next to the bill: the work at API list prices, divided by the seat cost. My piece on Techstrong.ai: techstrong.ai/features/your-…
5
1
11
145
Mo. Abubakar Lala retweeted
All roads in marketing lead to paid ads eventually
232
330
110
3,132
142,301
Pakistani freelancers earned a record $1.76bn in FY2025-26. Most of it still comes from selling hours, the model AI disrupts first. The hour is getting cheaper. The person who understands the client is not. My piece in The News: thenews.pk/magazine/us/14481…
1
2
15
157
Mo. Abubakar Lala retweeted
SEO is dead. Again. This is the 47th time this year. Some guy with a ChatGPT login and a ring light said so, so it must be true. Ignore the fact that our clients are still getting leads from Google every single day. Ignore the fact that AI tools literally pull their answers from websites that rank. Because a man on LinkedIn wrote “SEO is dead” with a skull emoji, we should all pack up and go home. RIP SEO (1997 - every Tuesday) 🪦
18
2
3
38
6,668
Mo. Abubakar Lala retweeted
95% is you 5% is prayer _@theablelala
1
1
1
10
One model, one pass, is how most AI work gets made. My Claude Code skill splits a job across Claude, GPT and Gemini, then a reviewer that has seen only my rules checks the result. It fixes what it finds. Follow, like and comment ORCHESTRATE, and I'll DM it to you.
1
2
11
108
90% accuracy per step × 5 steps = 59%. That's why most AI agents fall apart on real work. I'm giving away the starter kit I run my agents on: 3 layers, job-role folders, one instruction file every AI tool reads. Follow, like and comment AGENTS, and I'll DM it to you.
3
3
11
82
Every prompt I write has 3 parts: Task: first this, then this, then this. Output: the exact result. A file, a post, a page, a PDF. Purpose: the real problem, in detail. Most people skip the third. It matters most. Follow, like and comment PROMPT, and I'll DM you the template.
1
5
56
SEO checklists are guesses. The pages already ranking in your town are the answer key. My Rulesourcing kit for Claude Code counts what the top 10 share and writes rules you can pass or fail. Re-run it monthly. Follow, like and comment RULES, and I'll DM it to you.
4
48
Your AI invoice shows what you paid. Seat Yield shows what you got. A free script: it reads your Claude Code logs, prices the work at API rates and divides by your seat cost. One number per person, one for the team. Follow, like and comment YIELD, and I'll DM it to you.
1
26
Motion graphics are just code now. One HTML file draws frame t. Chrome screenshots each frame. ffmpeg makes the MP4. Free kit: the renderer, a sample reel, my motion playbook and a Claude Code skill that builds reels. Follow, like and comment MOTION, and I'll DM it to you.
63
Stop asking one AI for a second opinion. My Claude Code skill asks Claude, GPT and Gemini the same question, blind. Then it shows where they agree, who disagrees and how sure the answer is. Follow, like and comment PANEL, and I'll DM it to you.
2
45
Your AI agent forgets everything you corrected yesterday. Mine doesn't. One fact per file, one index it reads every session, and every correction saved with the reason. No database, no plugin. Just files. Follow, like and comment MEMORY, and I'll DM you the template.
16
Most AI embarrassments are one confident, made-up number. Nothing my agents make reaches a client until a named person signs it off. The agent runs a QA checklist first, so review takes minutes. Follow, like and comment GATE, and I'll DM you the checklist and sign-off kit.
8
I let an AI agent run my browser: profiles, listings, forms. It has six hard stops where it must hand back to me: payments, CAPTCHAs, phone or ID checks, passwords, account warnings, anything legal. Follow, like and comment STOP, and I'll DM you the full rule set.
14