What is the difference between prompt engineering and context engineering?

Prompt engineering crafts the instruction: wording, structure, examples. Context engineering designs everything the model knows when it works: retrieved documents, memory, codified business knowledge, assembled per moment. The prompt asks; the context informs. Reliable business AI is mostly won at the context level, with good prompts embedded in it.

What is loop engineering compared to prompt engineering?

Prompt engineering optimises one instruction you give by hand; loop engineering designs the autonomous system that decides what to prompt, when, and whether the result is acceptable, running on a schedule toward a defined goal. The prompts do not disappear; they get embedded in a system that issues them without you.

Is prompt engineering dead in 2026?

No: it moved indoors. Every loop and every context pipeline executes prompts, and their quality still shapes output. What ended is prompting as the whole job: the leverage moved up to what the AI knows (context) and what system drives it (loops), with instruction-craft embedded in both.

What is context engineering in simple terms?

Managing the AI's desk: making sure the right documents, memory and business knowledge are in front of the model at the moment it reasons. RAG, agent memory and codified knowledge files are its main tools. Most wrong answers about your business are desk problems, not model problems.

Which should my business learn first?

In order: prompt basics for everyone (a day), context engineering for anything answering about your business (your first real project), loop engineering once one governed agent runs well (the compounding step). Skipping levels is the expensive path: loops amplify whatever quality the lower levels feed them.

Are these formal disciplines or just buzzwords?

The labels are recent and will churn; the structure is real and durable: instruction quality, knowledge quality, system quality, each capping the next. Treat the terms as handles on that structure, and evaluate any new coinage the same way: which level is it naming, and is that your bottleneck?

What is harness engineering and where does it fit?

A sibling term from the developer world: designing the environment one agent runs inside (its tools, permissions, checks and working setup), sitting between context engineering and loop engineering in the stack: the harness equips the worker, the loop schedules and drives it. Harness work maps to the tool calling and governance layers; the kitchen version is fitting out the station itself. Expect the term to appear in vendor material; place it on the stack and it stops being confusing.

Can one person in a small business cover all three levels?

Realistically, yes, and many do: a technically curious ops manager can write the brief-pack (level 1), own the document hygiene and grounding relationship (level 2), and operate a partner-built loop (level 3). What one person should not do is build level 3 alone without the lower levels' evidence behind them; the levels are also a seniority ladder for the automation itself, and promotion by results applies to systems exactly as to staff.

What does agentic AI mean in simple terms?

AI that works toward a goal rather than answering a single prompt: it plans steps, uses tools, remembers context, checks results and adjusts, escalating to humans at the edges of its authority. Cruise control follows a setting; a chauffeur handles a destination. Agentic AI is the chauffeur.

What is the difference between agentic AI and generative AI?

Generative AI produces content when asked: text, images, code. Agentic AI uses that generative capability as a reasoning engine inside a larger loop of planning, tool use and feedback to complete tasks. Every agentic system contains generative AI; most generative AI use is not agentic.

What is the difference between agentic AI and an AI agent?

Agentic AI is the pattern; an AI agent is one system built on it. In practice the terms travel together: an agent is the worker, agentic describes how it works, and multi-agent systems are teams of them.

Is ChatGPT agentic AI?

As a chat window, no: it responds to prompts. Given tools, memory and goals, which modern AI platforms increasingly support, the same model becomes the reasoning engine of an agentic system. Agentic is a property of the architecture around the model, not of the model itself.

What are examples of agentic AI in small businesses?

Returns processing within policy limits, invoice chasing matched to customer history, CRM records maintained from real emails, supplier research and comparison, support tickets resolved end to end with escalation paths. The common shape: one bounded goal, scoped tools, humans on the exceptions.

Should my business adopt agentic AI now or wait?

Adopt the sequence, not the headline: prove value with assistants and simple automation first, then deploy one bounded agent on a measured workflow, then expand on the plumbing you have paid for. Our pillar guide, Agentic AI vs Workflow Automation, runs this decision with real ROI maths, and the 90-Day Roadmap turns it into a calendar.

What is the difference between agentic AI and RPA (robotic process automation)?

RPA scripts a computer to click and type like a human would: fixed steps, brittle when screens change, no judgment. Agentic AI reasons toward goals and adapts. They are not rivals in practice: RPA often serves as hands for legacy systems without APIs, while the agentic layer does the deciding. The distinction to hold onto: RPA follows steps a human recorded; agents choose their own.

Who are the main platforms and frameworks for agentic AI?

The model providers, Anthropic, OpenAI and Google, supply the reasoning engines; frameworks like LangGraph, CrewAI and the OpenAI Agents SDK structure agents and teams; automation platforms such as n8n, Zapier and Make now ship visual agent builders; and MCP standardises tool connections across all of them. Our AI Automation Stack guide tours every layer, and our n8n vs Zapier vs Make comparison covers the platform tier. The strategic point: the architecture pattern is stable even as the logos rotate, so learn the pattern.

What is an agent loop in AI?

The repeating cycle that makes AI autonomous: assemble context, reason and choose one action, execute it, observe the result, check whether the goal is met or a boundary hit, and repeat. Every agentic system runs this rhythm; quality differences between agents are mostly quality differences between their loops.

What is the ReAct pattern?

Reason plus Act: the 2022 research insight that interleaving a model's reasoning with real actions, think, act, observe, think again, grounds each thought in evidence rather than letting the model hallucinate past reality. It is the ancestor of most modern agent loops and the default rhythm of single agents.

How many cycles does an agent task take?

Typically a handful to a few dozen: a worked example can run eight cycles in about forty seconds. Cycle counts balloon when context is poorly assembled or stopping conditions are missing, which is a design smell rather than a task property; budgets per task keep the rhythm honest.

What is the difference between an agent loop and loop engineering?

The agent loop is the inner rhythm: one task, cycled to completion. Loop engineering designs the outer system: a scheduled, goal-driven loop that finds work, dispatches inner loops, verifies, remembers and continues indefinitely. Inner loops existed all along; the outer discipline is what got named in 2026.

Why do agents get stuck repeating the same steps?

Circling: the loop lacks loop-detection, step budgets or a route-level plan, so locally sensible cycles repeat at machine speed. Standard fixes are budgets, detection, an upfront plan and replan triggers.

Does a better AI model mean a better loop?

It helps the reasoning step and nothing else. Context assembly, observation honesty, done-checks, budgets and logs are engineering, not model quality, and they are where most real-world agent differences live. Feed and check the loop before paying for a bigger brain.

What is a stopping condition in an AI loop?

An externally checkable rule that ends or pauses a loop: the goal verifiably met (done-check), resources spent (budgets), a threshold crossed (boundaries routing to humans), or a failure pattern detected (circuit breakers). The defining property: checkable outside the AI's own opinion of itself.

Why can't the AI just decide when it's finished?

Because a model's done is a fluent claim, produced by the same mechanism as its correct and incorrect answers alike. Trustworthy done-checks use mechanical verification or a separate checker model, the maker-checker split applied to completion itself, which is exactly how the leading coding-agent tools implement their run-until-done features.

How do I prevent runaway AI costs in loops?

Four layers: hard budgets per task and per run (cycles, tokens, actions, time), cost-velocity alerts (spend rate versus baseline), context discipline so cycles do not bloat, and discovery sanity bounds so a data hiccup cannot feed the loop an ocean. Budgets are the layer that works when everything else surprises you.

What is a circuit breaker in AI automation?

A stop triggered by failure patterns rather than single events: repeated check failures, recurring tool errors, quality collapse or abnormal spend velocity, pausing the loop and alerting a named owner. Borrowed from electrical engineering for the same reason: patterns of fault deserve automatic disconnection.

What should be in a loop's stop-sheet?

One page: the verifiable done-condition per item and per run, the four budgets (cycles, tokens, actions, time), the boundary thresholds routing items to human approval, the circuit-breaker rules, and the named owner of halts with a one-line runbook. If it does not fit a page, the loop's scope is probably too broad.

How do stopping conditions relate to human approval gates?

Gates are one of the four stop types (boundaries): conditions that route an item to a person rather than ending the run. The gate-map design from our human-approval guide slots directly into the stop-sheet; loops simply give the gates a schedule and make their absence more expensive.

What is maker-checker in AI loops?

The structural rule that the agent producing work is never the one verifying it: a second agent, differently instructed and ideally a different model, reviews every item against explicit criteria before the loop proceeds, and completion itself is judged by something other than the worker. In unattended loops this split is not best practice; it is the component that replaces the absent human.

How do you verify AI output when nobody is watching?

Three built layers: mechanical checks on every testable fact at action time, an independent checker agent per work item, and verified done-conditions per run, all writing to a catch log, backed by a permanent human sampling habit that measures the escape rate. Designed before walking away, or it does not exist.

The record of everything verification intercepted: mechanical failures, checker rejections, done-check disputes, with reasons. It proves the checks are alive, supplies the tuning data since catches cluster around real causes, and answers audits. An active loop with an empty catch log has dead checks, not a perfect agent.

Can two AIs checking each other really be trusted?

More than one AI trusted alone, less than the marketing implies: different model families reduce shared blind spots, mechanical verification outranks both, and sampled human review remains the epistemic floor. The design goal is not certainty; it is a known, small, monitored escape rate with bounded consequences.

How much does loop verification cost?

Mechanical checks are effectively free; a checker pass adds a modest model call per item, pennies; the human sample costs minutes on a cadence. Against unattended mistakes at machine speed, it is the most lopsided trade in the discipline, the reflection economics compounded by schedule.

What happens when the checker and maker disagree repeatedly?

Revision caps end the ping-pong, two redrafts then the human queue, and persistent disagreement is treated as signal, not friction: work the loop cannot satisfy itself about is precisely what a person should see. A spike in disagreements trips the quality breaker and pauses the run, because it usually means the world changed upstream.

Who in a small business should own loop verification?

The same person who owns the loop: usually whoever held the responsibility before it was automated. Their verification duties are concrete and bounded: the daily exception queue (minutes), the weekly catch-log read (fifteen minutes), the monthly sample, and the quarterly checklist review. It is a management role over a tireless report, not a technical one, and the businesses that thrive at this treat it as a named part of a real job description rather than a hope.

What is an AI agent in simple terms?

Software that completes tasks toward a goal: a reasoning engine (brain) plans the steps, tools (hands) act on real systems, memory (notebook) holds context, a feedback loop (eyes) checks results, and boundaries (job description) keep it inside its authority with humans on the exceptions.

How is an AI agent different from ChatGPT?

ChatGPT in a chat window is a brain alone: brilliant conversation, no hands, no persistent job. An agent wraps that same class of brain with tools, memory, feedback and boundaries so it can actually complete work. The model is one component of five, not the product.

What are the main components of an AI agent?

Five: the reasoning engine (an LLM), tool connections to real systems, layered memory, a feedback loop that observes and adjusts, and explicit boundaries with escalation paths. Production quality lives mostly in the last four, which demos rarely show.

Do AI agents make mistakes?

Yes, and well-built ones are designed around that fact: grounded memory reduces wrong facts, feedback loops catch failed steps, boundaries cap the damage of wrong actions, and human gates hold the consequential calls. The realistic promise is not zero errors; it is small, caught, reversible, explainable ones.

How much does an AI agent cost to build?

For SMB single-workflow agents: typically £8,000 to £25,000 to build plus running and maintenance, with tool integrations usually the largest line. The anatomy explains the spread: more tools, richer memory and stricter governance all add parts.

Can I build an AI agent without coding?

Increasingly yes, for moderate cases: platforms like n8n ship visual agent components, a brain node, tool nodes, and memory. The anatomy stays identical; only the assembly method changes. Complex, regulated or high-volume agents still justify custom builds.

How long does it take to build an AI agent?

For a bounded single-workflow agent with the full anatomy: typically four to eight weeks, of which the brain is days and everything else is the schedule: tool integrations (longest, and dependent on your systems' APIs), memory design, boundary definition and the shadow-testing period that earns launch. Timelines quoted in days are quoting a demo; timelines quoted in quarters, for one workflow, deserve a second opinion.

Do I need a different agent for every task?

Broadly, one agent per coherent job description, which usually means per workflow: the returns agent, the invoice-chasing agent, the intake agent. They share plumbing (connections, memory infrastructure, governance patterns), which is why the second agent costs roughly half the first. Stretching one agent across unrelated jobs produces fuzzy scope and predictable failure; hiring a team of specialists is the multi-agent approach instead.

Agent Memory Explained: How AI Remembers Your Business (and Why It Forgets)

Aug 22

Written By Alpa Parmar

AI agents are amnesiacs by default. Agent memory fixes that in layers: working memory, long-term memory and retrieved knowledge. How each works, what goes wrong, and the design questions buyers should ask.

Here is the strangest fact about the technology behind every AI agent: the brain has no memory.

A large language model, for all its brilliance, wakes up fresh with every request. It does not remember your last conversation, your customer's name, or the thing it did thirty seconds ago, unless something outside the model deliberately reminds it. Every appearance of memory in every AI product you have used is engineering: systems built around the brain to catch, store and re-supply what matters.

For a business deploying agents, this is not trivia; it is the difference between an agent that feels like a competent colleague and one that feels like a goldfish with a good vocabulary. Memory design decides whether your agent knows the customer already emailed twice this week, whether it repeats yesterday's mistake, and whether it quotes this month's prices or last year's. So let's open this particular organ up properly: the three layers of agent memory, how each works, where each fails, and what to ask anyone building yours.

(This is one chapter of our agent anatomy series; the full skeleton is in [What Is an AI Agent?], and the concept overview in [What Is Agentic AI?].)

The Amnesiac Genius: Why Memory Must Be Built

One mental model makes everything else in this article click.

The analogy: imagine hiring the world's most capable consultant, with one condition: total amnesia between meetings, and even within a long meeting, notes slide off the table when it gets too full. Brilliant, but structurally forgetful. How would you make this consultant useful? Exactly the way the industry does: a notepad in the meeting (working memory), a filing system between meetings (long-term memory), and a research library they can consult on demand (retrieved knowledge). The consultant never stops being amnesiac; the system around them stops it mattering.

Every agent memory architecture is those three layers, so let's take them one at a time.

Layer 1: Working Memory. The Notepad in the Meeting

What it is: the context window: everything the model can "see" right now: the conversation so far, the task instructions, the data from recent tool calls. This is the only memory the brain natively has, and it is bounded: big (modern models hold hundreds of pages), but finite, and everything in it vanishes when the session ends.

How it fails, in two familiar ways. First, overflow: long tasks generate more history than the window holds, so something must be dropped or summarised, and badly chosen summaries lose the detail that mattered (the classic "it forgot what we agreed an hour ago"). Second, dilution: even within limits, models attend less sharply to a window stuffed with irrelevance; a focused notepad beats a hoarder's one.

The design work you are paying for: deciding what stays verbatim, what gets summarised, and what gets pushed out to longer-term storage, continuously, mid-task. Good builds manage the notepad like a chief of staff manages a leader's briefing: ruthless relevance.

Layer 2: Long-Term Memory. The Filing System Between Meetings

What it is: everything the agent should know across sessions, stored outside the model and re-supplied when relevant: customer records and histories, past interactions, decisions made, preferences learned, tasks completed. In practice this lives in ordinary, battle-tested databases (PostgreSQL being the workhorse, as our [AI automation stack] guide covers): exact facts, exact records, queryable forever.

The two flavours worth knowing. Episodic memory: what happened ("this customer returned two jackets in March, we refunded £240 on approval"). And semantic memory: what is true ("this customer prefers email, is on the trade discount, flagged as VIP"). Good agents use both: the episode explains, the fact directs.

How it fails. Staleness: facts recorded once and never updated ("prefers phone" from 2024). Bloat: everything stored, nothing curated, so retrieval drowns in noise. And the subtle one, wrong abstraction: storing raw transcripts when the agent needed conclusions, or conclusions when it needed the verbatim quote. Memory curation (what to write down, in what form, when to update, when to forget) is unglamorous design work and it is most of the difference between agents that learn and agents that hoard.

Layer 3: Retrieved Knowledge. The Research Library

What it is: your business's documents (policies, prices, product specs, procedures) made searchable by meaning, so the agent can pull the right page at the right moment. This is RAG with its vector database, and it earned its own full article: [Vector Databases and RAG Explained]. The one-line refresher: the library does not live in the agent's head; it gets consulted, page by relevant page, which is why updating a document updates the agent instantly.

Why it is a separate layer. The filing system holds what happened with YOUR specific customers and tasks; the library holds what is TRUE about your business generally. Different update rhythms, different owners, different failure modes: the library fails through outdated documents (fix the handbook, not the agent), the filing system through poor curation.

How the Layers Work Together: One Email, Three Memories

A customer emails your support agent: "Hi, following up on my issue from last week."

Working memory holds the email and the agent's instructions. Long-term memory supplies the episode: last week's ticket, what was tried, the promised follow-up. The library supplies the policy: what the escalation path for this product is this quarter. The reasoning engine gets all three delivered onto its notepad, and replies like a colleague who was there last week, because the system around the amnesiac made it so.

Remove any layer and the failure is recognisable: without long-term memory, "please describe your issue" (rage); without the library, invented policy (risk); without working-memory discipline, losing the thread mid-conversation (distrust). When users say an agent "feels smart," they are usually complimenting its memory design, not its model.

Why Should a Business Buyer Care?

Because memory is the moat, not the model. Every competitor rents the same brains. What they cannot rent is your filing system and your library: the accumulated episodes, curated facts and current documents that make YOUR agent know YOUR business. Memory is where an agent deployment compounds value over time, which also means memory neglect is where deployments quietly rot.

Because most "AI got it wrong" complaints are memory complaints. Wrong price: stale library. Repeated question: missing episodic memory. Forgot mid-task: notepad overflow. Diagnosing by layer turns mysteries into maintenance tickets, and the fix is usually a document or a curation rule, not a new model.

Because privacy lives here. Memory is where customer data persists, so this layer carries the data-protection weight: what is stored, for how long, who can query it, what gets forgotten on request. The [governance guide]'s least-privilege and audit principles apply to memory reads and writes exactly as to actions, and "what does the agent remember about people?" belongs in every deployment review.

What Are the Limitations?

Forgetting must be designed too. Real businesses need deletion: customers who leave, data-retention limits, the right to be forgotten. A memory system with no forgetting policy is a compliance incident maturing quietly.

Memory can entrench errors. A wrong conclusion written to long-term memory becomes tomorrow's confident context. Good builds mark memories with provenance and confidence, and the [reflection] loop re-checks what matters.

Shared memory is a coordination problem. Once [multiple agents] read and write the same filing system, you inherit the classic issues: conflicting updates, one agent's error becoming everyone's belief. Same solution as human teams: clear ownership of records and review at the writes.

More memory is not better memory. Retrieval quality beats storage quantity everywhere in this field. The goldfish with a perfect filing system outperforms the hoarder with a warehouse.

The Memory Audit: Five Checks for a Running System

For businesses that already have an agent live, here is the half-day audit that finds the leaks before customers do.

Check 1: The staleness sweep. Pick ten facts the agent relies on (prices, policies, key customer attributes) and trace each to its source: is the document current, is the stored fact dated, when was it last verified? Anything without a "last updated" trail is a wrong answer waiting for its moment.

Check 2: The retrieval spot-test. Ask the agent ten real questions and inspect not just the answers but what was retrieved to produce them: the right pages? The current versions? A correct answer built on lucky retrieval fails next week; the retrieval, not the answer, is what you are auditing.

Check 3: The continuity test. Run a two-session scenario: raise an issue today, follow up "on it" tomorrow. Does the agent connect them, and is what it remembered appropriate (the conclusion, not necessarily the verbatim rant)? This one test exercises all three layers at once.

Check 4: The forgetting drill. Request deletion of a test customer's data, then probe: does any trace survive in long-term memory, logs, or retrievable summaries? If the deletion path has never been tested, it does not exist; auditors agree.

Check 5: The write review. Sample twenty recent long-term memory writes: are they curated conclusions or raw transcript hoarding? Correct? Attributed and dated? This sample predicts your memory quality a year out better than anything else, because writes are the compounding input.

Half a day, quarterly, owned by whoever owns the agent. Most businesses that run it the first time find two staleness leaks and one hoarding habit, all fixable in an afternoon, all invisible until a customer would have found them instead.

The Buyer's Memory Questions

Five questions for any vendor or builder:

What does the agent remember across sessions, and where does that live?
What gets written to long-term memory, who decided that rule, and how is it kept current?
Is business knowledge retrieved from OUR documents, with citations, and what is the update cadence when documents change?
What is the forgetting policy: retention limits, deletion requests, stale-fact review?
Show me a wrong answer traced to its memory source, and the fix.

Fluent answers mean the memory was designed. Pauses mean it accreted, and accreted memory is where next year's mysteries are being stored today.

Frequently Asked Questions

Do AI agents have memory built in? No. The underlying models are stateless: they forget everything between requests. All agent memory is engineered around the model in layers: a managed context window (working memory), databases for cross-session facts and history (long-term), and retrieval from your documents (knowledge). Quality varies exactly as much as that engineering does.

What is a context window in simple terms? The model's notepad: everything it can consider at once, from instructions to conversation history to fresh data. It is large but finite, and managing what deserves space on it, verbatim, summarised or archived, is a core part of agent design.

What is the difference between RAG and agent memory? RAG is one layer of it: retrieving relevant knowledge from your documents at question time. Full agent memory adds working-memory management and long-term episodic and factual storage about specific customers and tasks. RAG answers "what is true about the business"; long-term memory answers "what happened with this customer."

Why does my AI keep forgetting things mid-conversation? Notepad overflow: the conversation outgrew the context window and something got summarised away or dropped. Better builds summarise selectively and push details to retrievable storage rather than losing them. If it forgets across sessions, the issue is absent long-term memory instead.

Is agent memory a privacy risk? It is where the privacy work lives: memory persists customer data, so retention limits, access scoping, audit logging and deletion policies all attach here. Well-designed memory is auditable and forgettable by policy; undesigned memory is a data-protection finding waiting for its auditor.

Can an AI agent learn from its mistakes? Yes, through memory plus reflection: outcomes get recorded, corrections get written back as curated facts, and future retrievals supply the lesson. It is bookkeeping rather than mystical learning, which is good news: bookkeeping is reliable, inspectable and yours.

How much does agent memory infrastructure cost? Modestly, by AI standards: long-term memory runs on ordinary databases (PostgreSQL, often already in your stack), and retrieval infrastructure (embeddings plus a vector store, or pgvector inside the same database) adds pounds-per-month hosting at SMB scale, not enterprise line items. The real cost is design and curation time: deciding what gets written, kept and forgotten. Which is good news, because design is a one-off and curation is a habit, while infrastructure bills recur. The [pricing guide]'s tiers absorb memory comfortably within normal build budgets.

Should different agents share memory? Share the library (business knowledge should be consistent everywhere) and be deliberate about the filing system: shared customer records need ownership rules for writes, while task-specific episodic memory usually stays per-agent. The failure mode to avoid is every agent writing freely to shared memory, where one agent's mistaken conclusion becomes every agent's confident context: the [multi-agent] coordination rules apply to memory exactly as to actions.

The Takeaway

The brain is an amnesiac genius; memory is the system that stops it mattering. Notepad, filing system, library: manage the first ruthlessly, curate the second deliberately, keep the third current, and design the forgetting as carefully as the remembering.

Get memory right and your agent compounds into something competitors cannot rent: an AI that actually knows your business. Get it wrong and you own a fluent goldfish. The model was never the moat. The memory is.

Bots and Brand Works designs agent memory as carefully as agent actions: curated writes, cited retrieval, forgetting policies included. If your AI keeps "getting things wrong," send us three examples and we will tell you which memory layer is leaking, free.

Need Help Implementing AI?

Resources and Further Reading

Pillar article: Agentic AI vs Workflow Automation: The 2026 Enterprise Guide
Related: Vector Databases and RAG Explained
Related: What Is an AI Agent? The Anatomy
Also in this series: Agent Planning · Reflection · Tool Calling · Human Approval
PostgreSQL: Click Here . pgvector: Click Here
Anthropic: Building effective agents: Click Here

Loop EngineeringRecursive Goals

Alpa Parmar

AI Marketing Automation

https://www.botsandbrandworks.com